Research Policy 44 (2015) 18271843

Contents lists available at ScienceDirect

Research Policy
journal homepage: www.elsevier.com/locate/respol

What is an emerging technology?

Daniele Rotolo a,b, , Diana Hicks b , Ben R. Martin a,c
a
SPRU Science Policy Research Unit, University of Sussex, Brighton BN1 9SL, United Kingdom
b
School of Public Policy, Georgia Institute of Technology, Atlanta 30332-0345, United States
c
Centre for Science and Policy (CSAP) and Centre for Business Research, Judge Business School, University of Cambridge, Cambridge CB2 1QA,
United Kingdom

a r t i c l e i n f o a b s t r a c t

Article history: There is considerable and growing interest in the emergence of novel technologies, especially from the
Received 11 December 2014 policy-making perspective. Yet, as an area of study, emerging technologies lack key foundational ele-
Received in revised form 15 June 2015 ments, namely a consensus on what classies a technology as emergent and strong research designs
Accepted 16 June 2015
that operationalise central theoretical concepts. The present paper aims to ll this gap by developing a
Available online 9 August 2015
denition of emerging technologies and linking this conceptual effort with the development of a frame-
work for the operationalisation of technological emergence. The denition is developed by combining
Keywords:
a basic understanding of the term and in particular the concept of emergence with a review of key
Emerging technologies
Conceptualisation
innovation studies dealing with denitional issues of technological emergence. The resulting denition
Denition identies ve attributes that feature in the emergence of novel technologies. These are: (i) radical novelty,
Attributes of emergence (ii) relatively fast growth, (iii) coherence, (iv) prominent impact, and (v) uncertainty and ambiguity. The
Operationalisation framework for operationalising emerging technologies is then elaborated on the basis of the proposed
Detection and analysis attributes. To do so, we identify and review major empirical approaches (mainly in, although not limited
Framework to, the scientometric domain) for the detection and study of emerging technologies (these include indica-
Scientometrics tors and trend analysis, citation analysis, co-word analysis, overlay mapping, and combinations thereof)
Indicators
and elaborate on how these can be used to operationalise the different attributes of emergence.
Science and Technology Studies (STS)
2015 Elsevier B.V. All rights reserved.

1. Introduction
Emerging technologies have been the subject of much debate
in academic research and a central topic in policy discussions and
initiatives. Evidence of the increasing attention being paid to the
phenomenon of emerging technologies can be found in the grow-
ing number of publications dealing with the topic and news articles
mentioning emerging technologies (in their headlines or lead para-
graphs), as depicted in Fig. 1. Increasing policy interest in emerging
technologies, however, must be set against a literature where no
consensus has emerged as to what qualies a technology to be
emergent. Denitions proposed by a number of studies overlap,
but also point to different characteristics. For example, certain def-
initions emphasise the potential impact emerging technologies are
capable of exerting on the economy and society (e.g. Porter et al.,
2002), especially when they are of a more generic nature (Martin,
Corresponding author at: SPRU Science Policy Research Unit, University of
Sussex, Brighton BN1 9SL, United Kingdom. Tel.: +44 1273 872980.
E-mail addresses: d.rotolo@sussex.ac.uk (D. Rotolo),
diana.hicks@pubpolicy.gatech.edu (D. Hicks), b.martin@sussex.ac.uk (B.R. Martin).
1995), while others give great importance to the uncertainty asso-
ciated with the emergence process (e.g. Boon and Moors, 2008) or
to the characteristics of novelty and growth (e.g. Small et al., 2014).
The understanding of emerging technologies also depends on the
analysts perspective. An analyst may consider a technology emer-
gent because of its novelty and expected socio-economic impact,
while others may see the same technology as a natural extension
of an existing technology. Also, emerging technologies are often
grouped together under general labels (e.g. nanotechnology, syn-
thetic biology), when they might be better treated separately given
their different socio-technical features (e.g. technical difculties,
involved actors, applications, uncertainties).
The lack of consensus over denitions is matched by an eclec-
tic and ad hoc approach to measurement. A wide variety of
methodological approaches have been developed, especially by the
scientometric community, for the detection and analysis of emer-
gence in science and technology domains (e.g. Porter and Detampel,
1995; Boyack et al., 2014; Glnzel and Thijs, 2012). These methods,
favoured, because they take advantage of growing computational
power and large new datasets and allow one to work with more
sophisticated indicators and models, lack strong connections to
well thought out concepts that one is attempting to measure, a basic

http://dx.doi.org/10.1016/j.respol.2015.06.006
0048-7333/ 2015 Elsevier B.V. All rights reserved.
1828 D. Rotolo et al. / Research Policy 44 (2015) 18271843

3500
300
News articles
Publications in all disciplines
Publications in social sciences

3000
250

2500
200

Number of news articles

Number of publications

2000
150

1500
100

1000
50

500
0

0
1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
Year

Fig. 1. Publications (left axis) and news articles (right axis) including the variations of the term emerging technologies. Publications were retrieved by querying SCOPUS:
TITLE(emerg* technol*) OR TITLE(emergence of* technolog*) OR TITLE(techn* emergence) OR TITLE(emerg* scien* technol*). Publications in social sciences were dened
as those assigned to the SCOPUS categories Business, Management and Accounting, Decision Sciences, Economics, Econometrics and Finance, Multidisciplinary,
Psychology, and Social Sciences. News articles were identied by searching for emerg* near2 technolog* in article headlines and lead paragraphs as reported in FACTIVA.
From 1980 to 2013, the average yearly growth rates of the number of publications concerning emerging technologies in all disciplines and in social sciences have been of
12.5% and 23.8%, respectively. The total number of publications in SCOPUS has yearly grown on average by 4.9%.
Source: search performed by authors on SCOPUS and FACTIVA.

tenet of good research design. Often no denition of the central con-
cept of an emerging technology is provided. It is no surprise there-
fore that approaches to the detection and analysis of emergence
tend to differ greatly even with the use of the same or similar meth-
ods. The operationalisation of emergence is also in a state of ux.
It changes as new categorisations (e.g. new terms in institution-
alised vocabularies, new technological classes) are created within
databases. This, in turn, makes less clear the exact nature of the phe-
nomena that these scientometric methods enable us to examine.
These problems in the effort to understand emerging technolo-
gies limit the utility of the research and so may hamper resource
allocation and the development of regulations, which, in turn, have
a major role in supporting and shaping the directionality of tech-
nological emergence.
The present paper addresses both the conceptual and method-
ological gaps. We aim to elaborate a framework that links what is
conceptualised as emerging technologies with its measurement,
thus providing guidance to future research (e.g. development of
novel methods for the detection of emergence and analysis of its
characteristics) and to policy-making (e.g. resource allocation, reg-
ulation). To do so, we rst attempt to clarify the conceptualisation
of emerging technologies by integrating different conceptual con-
tributions on the topic into a more precise and coherent denition
of emerging technology. We begin with the denition of emer-
gence or emergent, which is the process of coming into being, or
of becoming important and prominent. This is then enriched and
contextualised with a review of major contributions to innovation
studies that have focused on technological emergence, highlight-
ing both their common and contradictory features. Conceptual
attempts to grapple with emergence in complex systems theory are
also discussed where relevant to the idea of emergent technology.
The result is the delineation of ve key attributes that qual-
ify a technology as emerging. These are: (i) radical novelty, (ii)
relatively fast growth, (iii) coherence, (iv) prominent impact, and
(v) uncertainty and ambiguity. Specically, we conceive of an
emerging technology as a radically novel and relatively fast growing
technology characterised by a certain degree of coherence persisting
over time and with the potential to exert a considerable impact on the
socio-economic domain(s) which is observed in terms of the compo-
sition of actors, institutions and patterns of interactions among those,
along with the associated knowledge production processes. Its most
prominent impact, however, lies in the future and so in the emergence
phase is still somewhat uncertain and ambiguous.
Second, the framework for operationalising emerging technolo-
gies is developed on the basis of the attributes we identied.
The scientometric literature forms the core of the methods dis-
cussed because, as mentioned, this eld has been remarkably active
in developing methodologies for the detection and analysis of
emergence in science and technology. The reviewed methods are
grouped into ve main categories: (i) indicators and trend analysis,
(ii) citation analysis (including direct citation and co-citation anal-
ysis, and bibliographic coupling), (iii) co-word analysis, (iv) overlay
mapping, and (v) hybrid approaches that combine two or more of
the above. Because scientometric techniques cannot address all the
attributes comprehensively, we also discuss approaches developed
in other elds.
The paper is organised as follows. The next section introduces
the concept of emergence and its various components. In Sec-
tion 3, these elements are integrated with key innovation studies
proposing denitions of technological emergence, and a denition
of emerging technologies is then elaborated. Section 4 reviews
methods to both detect and analyse emergence, and then examines
the use of those approaches to operationalise the proposed deni-
tion and the various attributes of emerging technologies. Section
5 discusses the limits of current methodologies for the detection
and analysis of emerging technologies and identies directions for
future research. Section 6 summarises the main conclusions of the
study.
 D. Rotolo et al. / Research Policy 44 (2015) 18271843
Table 1
Dictionary denitions of the concept of emergence.
Dictionary denition of emerge/emergent
the process of coming into being, or of becoming important and prominent (New Oxford American
Dictionary)
to become manifest: become known [. . .] (Merriam-Websters Collegiate Dictionary)
to rise up or come forth [. . .] to become evident [. . .] to come into existence (The American Heritage
Desk Dictionary and Thesaurus)
move out of something and become visible [. . .] come into existence or greater prominence [. . .] become
known [. . .] in the process of coming into being or prominence (Concise Oxford English Dictionary)
starting to exist or to become known [. . .] to appear by coming out of something or out from behind
something (Cambridge Dictionaries Online)
Source: search performed by authors on major English dictionaries.
2. The concept of emergence
The word emerge or emergent means the process of coming
into being, or of becoming important and prominent (New Oxford
American Dictionary) or to rise up or come forth [. . .] to become
evident [. . .] to come into existence (The American Heritage Desk
Dictionary and Thesaurus). Table 1 presents dictionary denitions
of emergent. The primary attribute of emergence is becoming
that is, coming into existence. Emergent is not a static property;
it is a label for a process. The endpoint of the process is variously
described as visible, evident, important or prominent. Thus, among
the dictionaries there is some disagreement as to whether acknowl-
edged existence is enough for emergence, or beyond that, a certain
level of prominence is needed in order to merit application of the
term emergence.
There is a second denition of emergent given the by the New
Oxford American Dictionary as: a property arising as an effect of
complex causes and not analysable simply as the sum of their
effects. An additional denition is: arising and existing only as a
phenomenon of independent parts working together, and not pre-
dictable on the basis of their properties. This concept of emergence
is used in the study of complex systems. It can be traced back
to the 19th Century in the proto-emergentism movement when
Lewes (1875) referred to emergent effects in chemical reactions
as those effects that cannot be reduced to the components of the
system, i.e. the effects for which it is not possible to trace all the
steps of the processes that produced them. Its application in the
study of the dynamics of complex systems in physics, mathemat-
ics, and computer science gave rise to other fundamental theories
and schools of thought such as complex adaptive system theory,
non-linear dynamical system theory, the synergetics school, and
far-from-equilibrium thermodynamics (see Goldstein, 1999).
A number of studies focusing on the denitional issue of emer-
gence were produced by scholars in complex system theory
see Table A1 in Appendix for an overview of the denitions of
emergence proposed by major studies in complex system theory.
Goldstein (1999), for example, dened emergence as the arising of
novel and coherent structures, patterns, and properties during the
process of self-organization in complex systems (1999, p. 49). An
ontological and epistemological denition of emergence is instead
developed by de Haan (2006). Ontological emergence is about the
properties of wholes compared to those of their parts, about sys-
tems having properties that their objects in isolation do not have
(2006, p. 294), while epistemological emergence it is about the
interactions between the objects that cause the coming into being
of those properties, in short the mechanisms producing novelty
(2006, p. 294).
Though research on complex systems may have a certain cachet
(and perhaps for this reason scholars of emerging technologies
sometimes attempt to work with the meaning of emergent as
conceived by the complex system approach), we maintain that
1829
Attributes
come into being; important;
prominent
become manifest; become
known
evident; come into existence
visible; prominent; become
known; come into being
become known; to appear
questions about emerging technologies are not fundamentally
about understanding the origins and the causal nature of full system
interaction; rather they are about uncertainty, novelty, identica-
tion at an early stage, and visibility and prominence. It is true that
some technologies in themselves may be complex systems in the
sense of exhibiting adaptation, self-organisation, and emergence,
an example being parts of materials science (Ivanova et al., 1998).
However, other technologies exhibit complicatedness rather than
complexity as dened in complex system theory for example,
engineering systems. These systems are designed for specic pur-
poses, but they do not adapt and self-organise to changes in the
environment (Ottino, 2004). It is also true that emerging tech-
nologies may arise from complex innovation systems (Katz, 2006),
but we would contend that in the phrase emerging technology,
emerging is generally understood in the standard sense, not the
complex system usage.
3. Dening emerging technologies
To further clarify what is meant by emerging technology, we
reviewed literature in innovation studies dealing with denitional
issues of emerging technologies. To identify relevant studies, we
searched for emerg* technolog*, tech* emergence, emergence
of* technolog*, or emerg* scien* technol* in publication titles
by querying SCOPUS (see the left-hand column of Table 2).1 We
restricted the search to the title eld to limit results to publications
primarily focused on emerging technologies. The search identied
a total of 2201 publications from 1971 to mid 2014.2 Within this
sample we selected those publications in social science domains,
thus reducing the sample to 501 records (see Fig. 1).
We then read the abstracts and accessed the full-text of these
studies where necessary both to identify additional documents
from the list of cited references and to exclude studies that are
not relevant to the scope of this paper. We found that about 50% of
the studies in the sample refer to a specic industrial context (e.g.
listing and discussing emerging technologies in a given industry)
or to the educational sector (e.g. emergence of novel technologies
to improve education and learning). These were deemed not rele-
vant to our study. The remaining studies were further examined
to identify those that develop or provide denitions of emerg-
ing technologies we searched for dening sentences within
the publication full-text by using the keywords listed above. This
1
The terminology of emerging technologies has become central to a number
of research traditions and especially to the scientometric, bibliometric and tech-
mining domains (cf. Avila-Robinson and Miyazaki, 2011), which, as discussed, have
been remarkably active in developing methods for the operationalisation of emer-
gence. In other words, emerging technologies have become a category of its own.
For this reason, we do not include epistemologically related terms, such as radical,
disruptive, discontinuous, nascent and breakthrough.
2
The search was performed on 13th May 2014.
1830 D. Rotolo et al. / Research Policy 44 (2015) 18271843

Table 2
Search strategies used to identify the set of relevant publications for the conceptualisation and operationalisation of emerging technologies.

Conceptualisation Operationalisation

Search terms emerg* technolog* emerg* technolog*

tech* emergence tech* emergence
emergence of* technolog* emergence of* technolog*
emerg* scien* technol* emerg* scien* technol*
emerg* topic*
emergence of* topic*

Field(s) of search Title Title, abstract, keywords

Focus Social sciences Scientometric journals: Journal of the Association for Information Science &Technology
(formerly the Journal of the American Society for Information Science &Technology), Journal
of Informetrics, Research Evaluation, Research Policy, Scientometrics, Technological
Forecasting &Social Change, Technology Analysis &Strategic Management

Number of studies 501 155

Source: authors elaboration as based on SCOPUS data.

led to a core set of 12 studies from science and technology (S&T)
policy studies, evolutionary economics, management, and sciento-
metrics that contributed to the conceptualisation of technological
emergence. These are listed with their denitions of emerging
technologies in Table 3. We analysed the textual content of the def-
initions reported in Table 3 to extract all the component concepts.
These were grouped into the attributes discussed below and used to
construct our denition of emerging technologies. Extracted con-
cepts excluded from our list of attributes will also be discussed.
The rst dening attribute of emerging technology, explicitly
included in two of the 12 core articles, is radicalnovelty: novelty
(or newness) (Small et al., 2014) may take the form of dis-
continuous innovations derived from radical innovations (Day
and Schoemaker, 2000) and may appear either in the method or

Table 3
Denitions of emerging technologies (studies are chronologically ordered).

Study Domain Denition (elaborated or adopted)

Martin (1995)
Day and Schoemaker (2000)
Porter et al. (2002)
Corrocher et al. (2003)
Hung and Chu (2006)
Boon and Moors (2008)
Srinivasan (2008)
Cozzens et al. (2010)
Stahl (2011)
Alexander et al. (2012)
Halaweh (2013)
Small et al. (2014)
S&T policy A generic emerging technology is dened [. . .] as a technology the exploitation of which will yield
benets for a wide range of sectors of the economy and/or society (p. 165)
Management [. . .] emerging technologies as science-based innovation that have the potential to create a new
industry or transform an existing ones. They include discontinuous innovations derived from radical
innovations [. . .] as well as more evolutionary technologies formed by the convergence of previously
separate research streams (p. 30)
S&T policy Emerging technologies are dened [. . .] as those that could exert much enhanced economic inuence
in the coming (roughly) 15-year horizon. (p. 189)
Evolutionary economics The emergence of a new technology is conceptualised [. . .] as an evolutionary process of technical,
institutional and social change, which occurs simultaneously at three levels: the level of individual
rms or research laboratories, the level of social and institutional context, and the level of the nature
and evolution of knowledge and the related technological regime. (p. 4)
S&T policy Emerging technologies are the core technologies, which have not yet demonstrated potential for
changing the basis of competition (p. 104)
S&T policy Emerging technologies are technologies in an early phase of development. This implies that several
aspects, such as the characteristics of the technology and its context of use or the conguration of the
actor network and their related roles are still uncertain and non-specic (p. 1915)
Management I conceptualize emerging technologies in terms of three broad subheads: their sources [. . .], their
characteristics [. . .] and their effects [. . .] Specically, I consider two aspects of the sources of
emerging technologies the relay race evolution of emerging technologies, and revolution by
application four characteristics of emerging technologies the clockspeed nature of emerging
technologies, convergence, dominant designs, and network effects and three effects of emerging
technologies shifting value chains, digitization of goods, and the shifting locus of innovation (from
within the rm to outside the rm). (pp. 633634)
S&T policy Emerging technology a technology that shows high potential but hasnt demonstrated its value or
settled down into any kind of consensus. (p. 364). The concepts reected in the denitions of
emerging technologies, however, can be summarised four-fold as follows: (1) fast recent growth; (2)
in the process of transition and/or change; (3) market or economic potential that is not exploited fully
yet; (4) increasingly science-based. (pp. 365366)
S&T policy [. . .] emerging technologies are dened as those technologies that have the potential to gain social
relevance within the next 10 to 15 years. This means that they are currently at an early stage of their
development process. At the same time, they have already moved beyond the purely conceptual stage.
[. . .] Despite this, these emerging technologies are not yet clearly dened. Their exact forms,
capabilities, constraints, and uses are still in ux (pp. 34)
S&T policy Technical emergence is the phase during which a concept or construct is adopted and iterated by [. . .]
members of an expert community of practice, resulting in a fundamental change in (or signicant
extension of) human understanding or capability. (p. 1289)
Management Characteristics of (IT) emerging technologies are uncertainty, network effect, unseen social and
ethical concerns, cost, limitation to particular countries, and a lack of investigation and research. (p.
108)
Scientometrics [. . .] there is nearly universal agreement on two properties associated with emergence novelty (or
newness) and growth. (p. 2)

Source: search performed by authors on SCOPUS and extended to cited references.

 D. Rotolo et al. / Research Policy 44 (2015) 18271843
the function of the technology. To achieve a new or a changed
purpose/function, emerging technologies build on different basic
principles (Arthur, 2007) (e.g. cars with an internal combustion
engine vs. an electric engine, cytology-based techniques vs. molec-
ular biology technologies). Novelty is not only a characteristic of
technologies deriving from technical revolutions, i.e. technologies
with relatively limited prior developments (e.g. DNA sequencing
technologies, molecular biology, nano-materials), but it may also
be generated by putting an existing technology to a new use. The
evolutionary theory of technological change views this as the spe-
ciation process of technology, that is the process of applying an
existing technology from one domain to another domain or niche
(Adner and Levinthal, 2002). The niche is characterised by a selec-
tion process that is different from the one where the technology
was initially applied. The niche specically may differ in terms of
adaptation (the needs of the niche) and abundance of resources. The
technology applied in the niche may adapt and then emerge as well
as potentially invading other domains including the initial domain
(giving rise to a revolution or a process of creative destruction).
This implies that evolutionary technology (those not characterised
by revolutionary technical developments) can also be radically
novel in domains of application different from those where the
technology was initially developed. Adner and Levinthal (2002)
provided a compelling example of the speciation process by repor-
ting on the evolution of wireless communication technology. This
technology was created for laboratory purposes, and specically for
the measurement of electromagnetic waves. Yet, it found numerous
subsequent applications. Wireless communication technology rst
enabled communication with locations (e.g. lighthouses) otherwise
not reachable with wired telegraphy. Then, applications expanded
to the transmission of voice (radiotelephony and broadcasting),
and, more recently, to data transmission (Wi-Fi). With each shift,
wireless communication technology appeared radically novel in
its new domain of application, although the technology itself had
existed since the early laboratory and telegraphy applications. The
evolutionary theory of technological change teaches us that radical
novelty may characterise innovations based on both revolutionary
and evolutionary inventions resulting from the speciation process.
However, the term evolutionary is also used to refer to incremen-
tal technological advances. To avoid ambiguity, we opted to use
the term radical novelty rather than revolutionary/evolutionary
and to contextualise it in relation to the domain(s) in which the
technology is arising.3
The second dening attribute of emerging technologies, iden-
tied by three of the 12 core articles is clockspeed nature
(Srinivasan, 2008) or fast growth (Cozzens et al., 2010), or at least
growth (Small et al., 2014). Growth may be observed across a
number of dimensions such as the number of actors involved (e.g.
scientists, universities, rms, users), public and private funding,
knowledge outputs produced (e.g. publications, patents), proto-
types, products and services, etc. As with the radical novelty
attribute, the fast growth of a technology needs to be contextu-
alised. A technology may grow rapidly in comparison with other
technologies in the same domain(s), therefore relativelyfastgrowth
may be a better term.
The third attribute of emerging technologies, identied by four
of the 12 core articles is coherence that persists over time. The core
articles variously describe this attribute as convergence of previ-
ously separated research streams (Day and Schoemaker, 2000),
3
The word novelty alone may also create ambiguity with regard to the types of
technologies we aim to include in our conceptualisation of emerging technologies.
Technologies of a more incremental nature, as derived from the improvement of
existing technologies, are somewhat novel. For the sake of conceptual clarity, we
therefore prefer to add the attribute radical to the word novelty.
1831
convergence in technologies (Srinivasan, 2008), and technolo-
gies that have already moved beyond the purely conceptual stage
(Stahl, 2011). Alexander et al. (2012) point instead to the role of
an expert community of practice, which adopts and iterates the
concepts or constructs underlying the particular emerging tech-
nology. The concept of a community of practice suggests that both
a number of people and a professional connection between those
people are necessary. Coming together, intertwining and staying
together are all entailed in coherence. Coherence refers to internal
characteristics of a group such as sticking together, being united,
logical interconnection and congruity. The status of external rela-
tions is also important. The emerging technology must detach itself
from its technological parents to some degree to merit a separate
identity. Furthermore, it must stay detached for some period of
time to be seen as self-sustaining (Glnzel and Thijs, 2012). As we
stated above, emergence is a process and coherence, detachment
and identity do not characterise a nal state, but are always in the
process of realisation, presenting challenging issues of boundary
delineation and classication. Perspective matters since an analyst
may see an exciting emerging technology about to make a major
economic impact in something a scientist sees as long past the
exciting emerging phase.
The fourth dening attribute of emerging technologies, identi-
ed by nine of the 12 core articles is to yield benets for a wide
range of sectors (Martin, 1995), create new industry or trans-
form existing ones (Day and Schoemaker, 2000), exert much
enhanced economic inuence (Porter et al., 2002), or change the
basis of competition (Hung and Chu, 2006). Corrocher et al. (2003)
also point to the pervasiveness of the impact that the emerg-
ing technology may exert by crosscutting multiple levels of the
socio-economic system, i.e. organisations and institutions, as well
as knowledge production processes and technological regimes.
Accordingly, we identify prominentimpact as another key attribute
of emerging technologies. Most of the core articles conceived the
prominent impact of emerging technologies as exerted on the
entire socio-economic system. In this usage the concept of emerg-
ing technologies becomes very close to that of general purpose
technologies and so excludes technologies prominent within a
specic domain. We wish to include relatively smaller scale promi-
nence in our denition. For example, a diagnostic technology may
emerge and signicantly reshape the clinical practices associated
with a given disease, profoundly affecting one disease domain but
not others. In other words, our denition allows for prominent
impact with narrow scope (emergence in one or a few domains),
as well as wide-ranging impact across domains and potentially the
entire socio-economic system (e.g. ICT and molecular biology). Such
a perspective suggests, as with the attributes of radical novelty and
relatively fast growth, the importance of contextualising the promi-
nent impact of the observed technology within the domain(s) from
which the technology emerges.
The nal dening attribute of emerging technologies, identied
in seven of the 12 core articles is that the prominent impact of
emerging technologies lies somewhere in the future the tech-
nology is not nished. Thus, uncertainty features in the emergence
process. The non-linear and multi-factor nature of emergence pro-
vides emergence with a certain degree of autonomy, which in turn
makes predicting a difcult task (de Haan, 2006; Mitchel, 2007).
As a consequence, knowledge of the probabilities associated with
each possible outcome (e.g. potential applications of the technol-
ogy, nancial support for its development, standards, production
costs) may be particularly problematic (Stirling, 2007). Core articles
expressed this attribute in terms of the potential that emerging
technologies have for changing the existing ways of doing things
(e.g. Boon and Moors, 2008; Hung and Chu, 2006; Stahl, 2011).
However, these denitions seem not to disentangle explic-
itly another important aspect of emergence from the concept of
1832 D. Rotolo et al. / Research Policy 44 (2015) 18271843

uncertainty. This is ambiguity. Ambiguity arises because proposed

(2014)
applications are still malleable, uid and in some cases contradic-

Small
et al.
tory, i.e. even the knowledge of possible outcomes of emergence

x
x
is incomplete. A variety of possible outcomes may occur because
social groups encountered during emergence hold diverging val-

Halaweh
ues and ascribe different meanings to the technology (Mitchel,

(2013)
2007). It is worth noting that uncertainty and ambiguity are, how-

x
ever, not mutually exclusive (Stirling, 2007). These are not discrete
conditions. A continuum exists as dened by the extent to which

Alexander
knowledge of possible outcomes and likelihood for each outcome

(2012)
is incomplete. For example, it may be problematic evaluating the

et al.
probabilities associated with known possible outcomes, but at the

x
x
same time there may also be a lack of knowledge of other possible
outcomes such as unintended/undesirable consequences deriving
from the (potentially uncontrolled) use of the technology. Uncer-

(2011)
Stahl
tainty and ambiguity are key starting concepts for a wide variety

x
x
x
of science and technology studies (STS) focusing on the role of the
expectations in technological emergence (e.g. van Lente and Rip,
1998).

Cozzens

(2010)
The studies reviewed here introduced various additional

et al.
concepts such as the science-based-ness, network effects, and

x
x
early-stage development of emerging technologies. While the last

Srinivasan
of these seems to be implicit in the denition of emergence and the
key role of networks (of users adopting the technology) is certainly

(2008)
not a unique feature of emerging technologies, the association

x
x
x
with science-based-ness is less clear. The importance of science
(especially public science) for the development of industrial tech-
nologies is widely accepted on the basis of substantial evidence

Boon and
(e.g. Narin et al., 1997). However, even today not all technological

(2008)
Moors
revolutions may depend on breakthrough advances in science. In
certain domains, a technology can be developed without the need

x
for deep scientic understanding of how the phenomenon under-
lying it works it is possible to know how to produce an effect

Chu (2006)
without knowing how an effect is produced (Nightingale, 2014, p. Hung and
4). For example, Vincenti (1984) provided evidence of this in the
case of the construction of airplanes in the 1930s. The different

x
x
parts of an airplane were initially joined using rivets with dome-
shaped heads. These types of rivets, however, caused resistance
Corrocher

to the air, thus reducing the aerodynamic efciency of the plane.

(2003)

As other dimensions of airplane performance were improving (e.g.

et al.

speed), the aerodynamic efciency became increasingly relevant.

x
Innovation studies dening emerging technologies

The dome-shaped rivets were therefore replaced with rivets ush

with the surface of the airplane. This was a major improvement for
the aerodynamics of airplanes in 1930s, but it required no major
(2002)
Porter
et al.

scientic breakthrough.4 A more recent example is the develop-

x
x

ment of smartphones which did not require major advancements

in science since most of the technologies used already existed
the integration of these technologies, and advances in design for
Schoemaker
Attributes of emergence and reviewed key innovation studies.

the creation of novel user interfaces instead provided the founda-

Day and

(2000)

tion of the innovation.5 For these reasons, science-based-ness does

not feature in our denition of emerging technologies.
x

x
x
x

In summary, as reported in Table 4, our review of innovation

studies identied ve main dening characteristics or attributes
of emerging technologies: (i) radical novelty, (ii) relatively fast
Martin
(1995)

growth, (iii) coherence, (iv) prominent impact, and (v) uncertainty

x

and ambiguity. Combining these attributes, we dene an emerging

Uncertainty and ambiguity

Source: authors elaboration.

Attribute of emergence

4
Other classical examples include prehistoric cave dwellers using re for cooking
Relatively fast growth

without any scientic understanding of it, the development of steam engines that
Prominent impact

predated the development of thermodynamics, or the Wright brothers testing ying

Radical novelty

devices before the eld of aerodynamics was established.

5
The innovation was architectural rather than modular according to the distinc-
Coherence

tion proposed by Henderson and Clark (1990). Also, smartphone technology can
Table 4

be considered as an example of emerging technology of an evolutionary nature.

As discussed above, the radical novelty of this technology is the result of existing
technologies converging in new domains of applications.
 D. Rotolo et al. / Research Policy 44 (2015) 18271843
technology as a radically novel and relatively fast growing technol-
ogy characterised by a certain degree of coherence persisting over time
and with the potential to exert a considerable impact on the socio-
economic domain(s) which is observed in terms of the composition
of actors, institutions and patterns of interactions among those, along
with the associated knowledge production processes. Its most promi-
nent impact, however, lies in the future and so in the emergence phase
is still somewhat uncertain and ambiguous.
It is reasonable to assume that the attributes of emergence range
from low to high levels. Nonetheless, to try and pin them down
to some absolute level is rather meaningless. As discussed, the
attributes of emergence (especially radical novelty and relatively
fast growth) provide an indication of emergence when they are
considered in the domain in which the given technology is arising
and therefore in relation to other technologies that may exist in
that domain. Most importantly, these attributes are likely to co-
evolve and assume very different levels over different periods of
emergence. In the early stage of emergence (pre-emergence), a
technology is likely to be characterised by high levels of radical
novelty as compared to other technologies in the domain in which
it is arising. However, the impact the technology can exert on that
domain is still relatively low. The technology has not yet gone
beyond the purely conceptual stage, multiple communities are
involved in its development, and the delineation of the boundary of
the technology is particularly problematic (i.e. low levels of coher-
ence). As a consequence, its growth is relatively slow or not yet
begun, and high levels of uncertainty and ambiguity are associated
with the future developments of the technology the technology
may not even emerge. The technology may then acquire a cer-
tain momentum. Some trajectories of development may have been
selected out and certain dimensions of performance prioritised and
improved. A community of practice may have also emerged. The
technology thus becomes more coherent. Its impact is also rela-
tively less uncertain and ambiguous, and the technology starts to
take off in terms of publications, patents, researchers, rms, pro-
totypes/products, etc. However, at the same time, it is likely that
the radical novelty of the technology will diminish other tech-
nologies that exploit different basic principles may be emerging as
well in the domain in which the considered technology is emerg-
ing. We conceived emergence as this phase where the attributes
of emergence are subject to dramatic change. Finally, impact and
growth may enter a stable or declining phase, the technology loses
its radical novelty, knowledge of the possible outcomes of the
technology becomes more complete (probabilities can be perhaps
assigned to outcomes), and the community of practice may become
well-established (e.g. regular conferences, dedicated journals). The
technology enters in a post-emergence period. In line with the
S-shaped patterns highlighted in early studies on the growth of
science (e.g. De Solla Price, 1963) and in technological adoption
literature (e.g. Manseld, 1961; Rogers, 1962), we stylised the
change in the levels of the attributes of emergence as following
an S curve (or more strictly, a reversed S curve in two of the ve
cases). This is qualitatively depicted in Fig. 2.
Dening emerging technology is, however, only half the battle.
If the denition is to be useful, we must show how the attributes
can be measured and thus how technologies can be classied as
emerging or not. In the next section, we link our denition to
the operationalisation of our denition of emerging technologies.
We rely mainly on scientometric techniques, bringing in other
approaches to ll certain gaps.
4. A framework for the operationalisation of emergence
Scientometric research has developed methods to detect emer-
gence in science and technology and is therefore central to
1833
Attribute
Relatively fast growth
Coherence
Prominent impact
Radical novelty
Uncertainty and ambiguity
Time
Pre-emergence Emergence Post-emergence
Fig. 2. Pre-emergence, emergence, and post-emergence: attributes and stylised
trends.
Source: authors elaboration.
operationalising our denition. From the vast literature that
touches on emerging technologies, we drew upon studies that offer
ideas on operationalising our ve attributes. We identied relevant
scientometric studies by including the term topic in the search
string we used to select research works dealing with denitional
issues of emerging technologies topic is often used in sciento-
metrics to refer to the emergence of a new set of research activities
in science and technology (e.g. Small et al., 2014; Glnzel and Thijs,
2012). The search was also extended to publication titles, abstracts
and keywords, but narrowed to journals mainly or to a signicant
extent oriented toward the publication of novel scientometric tech-
niques (see the right-hand column of Table 2). The search in SCOPUS
returned 155 publications.
The examination of cited references of these publications
enabled us to retrieve additional studies that were not cap-
tured with the search string, but are potentially relevant to for
our analysis. This increased the initial sample to 183 studies.
We then analysed these publications to identify studies that
were relevant to the operationalisation of the attributes of emer-
gence. This process led to a nal set of 55 publications,6 which
were then classied in terms of the methodological approach
adopted to detect or analyse emergence (e.g. indicators, citation
6
We excluded 76 studies that did not operationalise emergence (e.g. use of
emerging technologies as empirical context for various analyses, examination of
ethical issues associated with emerging technologies), three studies focused on the
review of scientometric methods for the analysis of emerging technologies, and
two studies elaborating document search strategies based on a modular lexical
approach. 33 studies that were concerned with Future-oriented Technology Analysis
(FTA) techniques (e.g. foresight, forecasting, roadmapping, Constructive Technol-
ogy Assessment (CTA)) were also not included in the review. While about 67% of
these do not rely on scientometrics, the remaining FTA studies in the sample pro-
pose frameworks for selecting, rather than identifying, emerging technologies, or
adopt conventional scientometric/bibliometric approaches, which will be instead
discussed with the review of the selected scientometric studies. FTA methods, how-
ever, remain crucial for more prospective analyses of emerging technologies and
decision-making on possible future scenarios (e.g. Porter et al., 2004; Irvine and
Martin, 1984; Ciarli et al., 2013). 14 STS studies included in the sample will be
instead referenced in our review and discussion when the operationalisation of the
attributes of emergence with the use of scientometric approaches is limited by a
lack of data or by the nature of the considered attribute. It is worth noting that
our search did not capture technometric studies (e.g. Grupp, 1994; Saviotti and
Metcalfe, 1984; Sahal, 1985). This research stream has been particularly important
for the measurement of technology and technological change. Nonetheless, techno-
metric models tend to rely on a variety of assumptions and often require data, the
collection of which can be particularly labour-intensive (e.g. extraction and coding
of data on the features of the considered technologies) (e.g. Coccia, 2005).
1834
Table 5
Methods for the detection and analysis of emergence in science and technology (studies are ordered by technique and publication year).
Method/study
Indicators and trends
Porter and Detampel (1995)
Kleinberg (2002)
Bengisu (2003)
Watts and Porter (2003)
Bettencourt et al. (2008)
Bettencourt et al. (2009)
Moed (2010)
Schiebel et al. (2010) and
Roche et al. (2010)
Guo et al. (2011)
Jrvenp et al. (2011)
Abercrombie et al. (2012)
Jun (2012) and Jun et al. (2014)
Avila-Robinson and Miyazaki
(2013b,a)
de Rassenfosse et al. (2013)
Ho et al. (2014)
Citations analysis
Direct citation
Seminal paper: Gareld et al.
(1964)
Kajikawa and Takeda (2008),
Kajikawa et al. (2008) and
Takeda and Kajikawa (2008)
Scharnhorst and Gareld
(2010)
Shibata et al. (2011)
Iwami et al. (2014)
Co-citation
Seminal paper: Small (1973)
Small (2006)
Cho and Shih (2011)
rdi et al. (2012)
Boyack et al. (2014)
Bibliographic coupling
Seminal paper: Kessler (1963)
Morris et al. (2003)
Kuusi and Meyer (2007)
Co-word analysis
Seminal paper: Callon et al.
(1983)
Lee (2008)
Ohniwa et al. (2010)
Yoon et al. (2011)
Furukawa et al. (2015)
Zhang et al. (2014)
Overlay mapping
Rafols et al. (2010)
Bornmann and Leydesdorff
(2011)
D. Rotolo et al. / Research Policy 44 (2015) 18271843
Data Operationalisation of emergence
Publications/patents Count of keywords in publication abstracts and trend analysis based on FisherPry curves
Publications/e-mails Burst of activity detected as state transitions of an innite-state automaton
Publications Positive slope of the line derived by regressing the number of publications on time and no
decrease of more than 10% or stability (no increase) in the last period or continuous decline in
the last three periods of observation
Publications Indicators of emergence: cohesion (based on cosine similarity between documents), entropy,
and F-measure
Publications Epidemic model to describe the increasing number of authors involved in an emerging eld
Publications Increasing densication (average number of edges per node), stable/decreasing diameter
(average path length between nodes), and increasing fractional count of edges in the largest
component of the co-authorship network
Publications Journals characterised by high values of Source Normalised Impact per Paper (SNIP) indicator
Publications Publication keywords initially labelled as unusual terms, by using tf-idf and Gini coefcient,
that subsequently become cross section terms, i.e. they diffuse in several research domains
Publications Indicators of emergence: frequency of keywords (ISI WoS keywords, authors keywords, and
MeSH terms), growing number of authors, and interdisciplinarity (based year-average
Rao-Stirling diversity index) of cited references
Mixed Absolute and cumulative count of the number of basic and applied research publications,
patents, and news
Mixed Normalised number of publications and citations, patents, and web news tted to a
polynomial function
News Normalised searching trafc (Google trends)
Publications/patents Overview of indicators to analyse emergence
Patents Count of the priority patent applications led by a countrys inventor, regardless of the patent
ofce in which the application is led
Publications Cumulative number of publications tted to a logistic curve
Publications
Publications Clusters of publications with the highest average publication year
Publications Historiographic approach combined with eld mobility of publications
Publications Clusters of publications with the highest values of betweenness centrality
Publications Publications (leading papers) with high values of in-degree (height), large variation of
in-degree between one year and the next year (slope), or large cumulative in-degree (area)
as dened on the basis of the yearly direct citation network
Publications
Publications Clusters with no continuing publications from the prior period
Patents Technological patent classes (IPC) that span structural holes in the co-citation network
Patents Clusters of patents present in a given time period and not in the previous period
Publications Yearly clustered publications of which references overlap less than 30% with references cited
by previous clusters
Publications
Publications Clusters of publications that cite more recent clusters of publications, namely emerging
research fronts
Patents Clusters of patents as source to identify guiding images (leitbild) of technological
development
Publications
Publications Clusters in the co-word network that show low values of degree, high betweenness, and low
closeness, i.e. those clusters that are more likely to turn into hub in the future.
Publications MeSH terms (clustered with co-word analysis) that are included in the top-5% by incremental
rate in a given year the increment rate for a MeSH term is dened as the number of time the
terms occurred at the time t, t + 1, and t + 2 out the number of times the term occurred at t 1,
t, t + 1, and t + 2
Patents Small and dense sub-networks in the invention property-function network
Publications Sessions of conferences in which previous sessions converge according to the average cosine
similarity (based on tf-idf-identied keywords) between the papers included in the sessions
Publications Combination of cluster analysis with term clumping and principal component analysis
Publications Overlays of publications projected on a basemap of ISI WoS subject categories linked by cosine
similarity of co-citations patterns between journals
Publications Overlays of publications on Google maps to identify cities publishing more than expected
 D. Rotolo et al. / Research Policy 44 (2015) 18271843 1835

Table 5 (Continued)

Method/study Data Operationalisation of emergence

Leydesdorff and Rafols (2011)
Leydesdorff et al. (2012)
Leydesdorff and Bornmann
(2012)
Leydesdorff et al. (2013)
Kay et al. (2014)
Leydesdorff et al. (2014)
Publications Overlays of publications and co-authorship networks on Google maps to trace collaboration
activity
Publications Overlays of publications projected on a basemap of MeSH terms linked by cosine similarity
(based on the co-occurrence of MeSH terms at the publication level)
Patents Overlays of patents on Google maps to identify cities patenting more than expected
Publications Overlays of publications projected on the basemap of journals linked by cosine similarity of
co-citations patterns between journals
Patents Overlays of patents projected on the basemap of 466 IPC classes linked by cosine similarity of
citing-to-cited relationships between classes the basemap is built by using patents included
in 2011 PATSTAT
Patents Overlays of patents projected on the basemap of 124 3-digit or 630 4-digit IPC classes linked
by cosine similarity based on co-citations between classes the basemap is built by using
patents granted at the United States Patent and Trademark Ofce (USPTO) from 1976 to 2011

Hybrid
Chen (2006): co-citation
analysis and burst detection
Leydesdorff et al. (1994):
co-citation analysis and
bibliographic coupling
Glnzel and Thijs (2012):
co-word, direct citation
analyses and bibliographic
coupling
Gustafsson et al. (2015):
co-occurrence of IPC classes
Small et al. (2014): direct and
co-citation analyses
Yan (2014): co-word analysis
and topic modelling
Chang and Breitzman (2009),
Breitzman and Thomas
(2015): direct citation and
co-citation analyses
Publications Trends in the bipartite network of research-front terms (burst detection) and intellectual base
articles the network includes three types of links: co-occurring research front terms,
co-cited intellectual base articles, and a research-front term citing an intellectual base article
Publications New journals that build on multiple existing areas, i.e. they load on multiple factors obtained
by the factor analysis of the matrix of the cited references, and have unique being cited
patterns, i.e. they are central tendency journals reporting highest load on a given factor as
obtained by the factor-analysis of the matrix of received citations
Publications Existing clusters with exceptional growth, completely new clusters with roots in other
clusters, and existing clusters with a topic shift
Patents Technological co-classication to identify clusters of patents and detect guiding images or
leitbild from patent full-text
Publications Clusters of publications that show high growth and are new both to the direct citation and
co-citation models
Publications Topics that are not a close variation of other topics, i.e. a topic i in the year t is emerging if no
predecessors are found and no other topics are transformed into topic i at t + 1
Patents Clusters of patents (co-citation clustering) that form around hot patents dened as those
patents that are highly cited (top 510%) by patents issued in the last two years and the
citations of which mostly come from patents issued in the last two years

Source: search performed by authors on SCOPUS and extended to publication cited references.

patterns between documents, co-occurrence of words in text), 4.1. Radical novelty

data sources used (e.g. publications, patents, news articles), and
proposed operationalisation of emergence. This information is
summarised in Table 5 where studies are grouped into ve
groups: (i) indicators and trend analysis studies that are mainly
based on document counts; (ii) citation analysis studies which
focus on examining citation patterns between documents; (iii) co-
word analysis studies that build on the co-occurrence of words
across document text; (iv) overlay mapping technique studies,
which use projections to position a given set of documents
within a wider or more global structure (e.g. a map of science);
and (v) hybrid studies that combine two or more of the above
approaches. Table 5 shows how denitions of emergence varied,
even within the same group of techniques, thus providing further
evidence of the low level of consensus on what constitutes emer-
gence.
Given the denitional weaknesses in the original studies, our
use of a particular study often varies from that of its authors.
We will briey introduce the major techniques and our interpre-
tation of the contribution they make to measuring attributes of
emerging technologies. For each attribute, we will rst describe
how it can be operationalised for contemporary and then for ret-
rospective cases of emerging technologies. When data scarcity
or the nature of the attribute of emergence limit the applica-
bility of scientometrics, we will discuss qualitative approaches.
The role of experts remains crucial for the validation of the
results obtained with the use of the techniques discussed below,
especially for qualitative approaches to the operationalisation of
emergence.
Emerging technologies are radically novel, i.e. they fulll a given
function by using a different basic principle as compared to what
was used before to achieve a similar purpose. Publications and
patents are of limited use in assessing radical novelty in contem-
porary technology. In contrast, news articles, editorials, review and
perspective articles in professional as well as academic journals
represent valuable sources, providing participant perspectives on
if and why a technology is viewed as radically novel. These docu-
ments may also provide an understanding of the basic principles
underpinning the examined technology.
In contrast, in retrospective analyses citation and co-word anal-
yses can be particularly effective for identifying radical novelty.
Relatively large amounts of data can be exploited to map the cog-
nitive networks of a knowledge domain over time. Citation analysis
builds on citation patterns among documents to generate a network
in which nodes are documents and links between nodes repre-
sent (i) a direct citation between two documents (direct citation
analysis) (Gareld et al., 1964), (ii) the extent to which two doc-
uments are cited by the same documents (co-citation analysis)
(Small, 1973), or (iii) to what extent two documents cite the same
set of documents (bibliographic coupling) (Kessler, 1963). Co-word
analysis instead exploits the text of documents to create a network
of keywords (or key phrases) that are linked according to the text to
which they co-occur across the set of selected documents (Callon
et al., 1983).
On the premise that clusters of documents or words in these
networks represent different knowledge areas of a domain or
1836 D. Rotolo et al. / Research Policy 44 (2015) 18271843
different literatures on which the domain builds, several studies
have considered the appearance of clusters not previously present
in the network as a signal of novelty (e.g. rdi et al., 2012; Kajikawa
and Takeda, 2008). Others dispute this interpretation. Given the
continuous evolution of science and technology, one is unlikely
to nd a cluster again in subsequent annual networks so the per-
centage of clusters that would qualify as newly appearing tends
to be relatively high. For this reason, additional criteria have been
suggested such as the appearance of new clusters that also link
otherwise weakly connected (e.g. betweenness centrality) clusters
(e.g. Shibata et al., 2011; Furukawa et al., 2015), that form around
documents that are highly cited by recent documents and the cita-
tions of which also are mostly from recent documents (Breitzman
and Thomas, 2015), or that cite more recent clusters as identi-
ed by the (Salton) similarity of their references (Morris et al.,
2003).
Small et al. (2014) have recently proposed a hybrid approach
based on a combination of direct citation and co-citation models as
applied to publication data. This approach is particularly focused on
the detection of novelty, which is dened in terms of clusters that
are new to the co-citation model that is, clusters with limited
overlap with the cited documents included in clusters in previ-
ous years (Boyack et al., 2014) as well as to a parallel direct
citation model. By combining bibliographic coupling, co-word anal-
ysis, and direct citation analysis, Glnzel and Thijs (2012) instead
dened novelty (namely emerging topics) as three cases of clus-
ters: those that show exceptional growth, those that are completely
new but with their roots in other clusters, or already existing ones
that exhibit a topic shift. Yan (2014) combined co-word analysis
with Natural Language Process (NLP) approaches (topic modelling).
Emergence, as reected in novelty, is then associated with the
appearance of topics that are not a close variation of other topics
calculated on the basis of the JensonShannon Divergence.7 Specif-
ically, a topic i appearing at time t is considered to be emerging
if it has no predecessors and none of the identied topics trans-
forms into topic i at t + 1. A different perspective is provided by
Scharnhorst and Gareld (2010) who extended the analysis of his-
toriographs (based on direct citations) to trace the extent to which
publications move across elds as they receive citations from new
elds (namely eld mobility). Assuming that these publications
are associated with a basic principle used for technological appli-
cations, this approach enables one to identify which elds may be
using a different knowledge base and thus in which elds radi-
cally novel technologies are potentially emerging. However, this
requires a priori knowledge of the basic principle and the set of
documents associated with it.
Research in scientometrics has also focused on the develop-
ment of techniques to expand the local (domain) perspective that
citation or text-based approaches may provide. This effort has gen-
erated a number of overlay mapping techniques (for an overview
see Rotolo et al., 2014), which in turn may be particularly well
suited to detecting radical novelty. The basic idea is to project a
given set of documents (e.g. publications associated with a research
domain) on a basemap through the use of an overlay. The basemap
can represent the global science structure at the level of the sci-
entic discipline (ISI Web of Science (WoS) subject categories) (e.g.
Rafols et al., 2010), journal (e.g. Leydesdorff et al., 2013), Medical
Subject Headings (MeSH) (Leydesdorff et al., 2012), or the techno-
logical structure at the level of patent classes (e.g. Kay et al., 2014;
7
The JensonShannon Divergence is a measure of similarity between empiri-
cally determined distributions (e.g. co-occurrence of words in documents) based on
Shannon entropy measures (for more details see Lin, 1991).
Leydesdorff et al., 2014).8 Once the set of documents (publications
or patents) associated with a given domain has been identied,
the projection of these documents over different time slices on
the global map of science or technology may reveal the increas-
ing involvement of new scientic or technological areas. This may
suggest that new knowledge areas are being accessed to conduct
research, and thus that potentially different basic principles are
drawn upon to achieve a given purpose.
Among the studies within the indicators and trends group of
techniques, Moed (2010) proposed the source normalised impact
per paper (SNIP) indicator for the evaluation of journals impact
and claims it is relevant for identifying emerging technologies. This
indicator is dened as the ratio between the journals raw impact
per paper (number of citations in the year of analysis to the jour-
nals papers published in the three previous years, divided by the
number of the journals papers in these three years) and the rela-
tive database citation potential in the subject eld covered by the
journal (mean number of 13-year-old references per paper citing
the journal and published in journals included in the considered
database divided by that for the median journal in the database).
Moed (2010) argued that the SNIP indicator, and specically high
values of this indicator, also provides information on the extent to
which a considered journal covers emerging topics. Given the focus
on recent citations and database coverage, the SNIP indicator is
clearly associated with the radical novelty attribute of emergence.
This indicator is, however, evaluated at the aggregate level of the
journal and journal-by-journal. It is therefore less clear whether
signals of radical novelty (i.e. relatively high values of SNIP) are
associated with one or multiple emerging topics the considered
journal may cover. In addition, the SNIP may not capture signals
of radical novelty in those instances of journals that cover few
emerging topics and therefore characterised by low values of SNIP.
All these techniques have various advantages and limitations.
The qualitative analysis of news articles, editorials, review and per-
spective articles, for example, may be effective for contemporary
analyses. However, the technical language used in these documents
may be an important barrier to a non-experts efforts to inde-
pendently assess radical novelty. The application of citation and
co-word analyses is strongly dependent on time. Data need to be
longitudinal in order to permit the tracing of cognitive dynamics
and associated changes in the knowledge structure. Co-word anal-
ysis and bibliographic coupling are, however, less sensitive to time
than direct citation and co-citation analyses and can be applied as
documents become available (e.g. Breitzman and Thomas, 2015).
Finally, overlay mapping provides a global perspective on emer-
gence for the assessment of radical novelty, but interpretation of
the resulting maps is mainly based on visual inspection.
4.2. Relatively fast growth
Emerging technologies show relatively fast growth rates com-
pared to non-emerging technologies. The assessment of this
attribute is particularly problematic for contemporary analyses.
Growth is not yet observed in terms of publications and patents, for
example, so scientometric indicators cannot be used. Early indica-
tions of growth may be revealed from the analysis of funding data,
big data, and altmetrics. This is an important research direction for
future studies on the operationalisation of the relatively fast growth
attribute, as we will discuss later in the paper.
8
The elements of the basemap are linked according to similarity based on the
co-occurrence of citations or, in the case of MeSH, the co-occurrence of terms. The
same approach can be used to project a sample of publications and patents onto geo-
graphical maps (e.g. Google maps) to reveal the most active cities and collaborative
activities (see Table 5).
 D. Rotolo et al. / Research Policy 44 (2015) 18271843
In the case of retrospective analyses, relatively fast growth is
perhaps the most frequently measured attribute of emergence in
scientometrics. Most studies assume rapid growth as a sine qua
non condition of emergence, and so a number of operationalisa-
tion approaches have been proposed. Indicators and trend analyses
based on the yearly or cumulative count of documents publi-
cations, patents or news articles, according to the nature of the
examined technology and the availability of data over a given
observation period are widely used. Documents are generally iden-
tied over time by using expert-dened keywords appearing in the
publication titles and abstracts (e.g. Porter and Detampel, 1995)
or by exploiting more institutionalised vocabularies such as the
MeSH classication in the case of publication counts in the biomed-
ical domain (e.g. Guo et al., 2011). With a focus on patent data,
de Rassenfosse et al. (2013) proposed counting the priority patent
applications led by a countrys inventor, regardless of the patent
ofce in which the application is led, as an indicator to identify
fast growth and therefore potential emerging technologies. How-
ever, yearly publication or patent counts are always dynamic, so the
problem becomes one of setting a criterion by which to distinguish
the signal from the noise, that is differentiating emerging technol-
ogy from other increasing trends. Some theoretical foundations are
needed to do this.
Rapid growth is also detected by tting the document count to a
function (e.g. forms of logistic function such as FisherPry curves).9
Bengisu (2003), for example, regressed the number of publications
over publication year and dened emerging technologies as those
technologies showing a positive slope and a decrease of less than
10% or stability (no increase) in the last period compared to the
previous one, or no continuous decline in the last three periods of
observation. Ho et al. (2014) instead tted the cumulative num-
ber of publications to a logistic curve, whereas Abercrombie et al.
(2012) extended the count of publications to patents, web news,
and commercial applications. Data were then normalised and t-
ted to a polynomial function for comparison a similar approach
is employed by Jrvenp et al. (2011) and Jun et al. (2014).
The number of documents is also used to detect bursts of activ-
ity, i.e. the appearance of a topic in a document stream. This relies
on the approach of Kleinberg (2002), who modelled the number of
publications and e-mails containing a given set of keywords as an
innite-state automaton, i.e. a self-operating virtual machine that
may assume a non-nite number of states and where the transi-
tion from one state to another is regulated by a transition function
(similarly to Markov models). The frequency of state transitions
with certain features identies bursts of activity, which are used
as a proxy for fast growth. The burst detection approach is com-
bined with co-citation analysis by Chen (2006) to build a bipartite
network10 of research-fronts linked with intellectual base arti-
cles. This network is then analysed in order to identify emerging
trends.
Schiebel et al. (2010) and Roche et al. (2010) proposed instead
an approach to emergence that is based on a diffusion model
(and diachronic cluster analysis to identify topics) that combines
9
FisherPry curves were developed to model technological substitution between
two competing technologies (Fisher and Pry, 1971). This family of curves is built
on the basis of three assumptions: (i) technological advancements are the result
of competitive substitutions of one method (technology) used to satisfy a given
need for another; (ii) the new technology completely replaces the old technology;
and (iii) the market share follows Pearls Law, i.e. the fractional rate of fractional
substitution of new for old is proportional to the remaining amount of the old left
to be substituted (Fisher and Pry, 1971, p. 75).
10
A bipartite network is a network in which nodes can be partitioned into two
distinct groups, N1 and N2 , and all the links connect one node from N1 with a node
from N2 , or vice versa (Wassermann and Faust, 1994).
1837
a modied tf-idf11 with the Gini coefcient to identify three stages:
unusual terms, established terms, and cross section terms.
Unusual terms are those that are rare in publications since they
describe a research discovery at a very early stage. When research
intensies, terms rst become more established in the original
domain and subsequently they may diffuse into other domains,
thus becoming cross section terms. Terms that change their clas-
sication (i.e. that show pathways) from unusual to cross section
terms from one period to another are characterised by rapid diffu-
sion and therefore relatively fast growth. This approach, however,
is highly dependent on the thresholds of the tf-idf and Gini coef-
cient selected to classify terms as well as on the duration of the
periods used to trace changes in the classication of terms.
Citation and co-word analyses can also be used to identify
the relatively rapid growth of a potential emerging technology.
Longitudinal analysis of the size of the clusters of documents or
words obtained with the application of these techniques can detect
knowledge areas that show rapid growth. For example, Ohniwa
et al. (2010) used co-word analysis to cluster MeSH terms. For each
MeSH term an increment rate was calculated in year t as the number
of times the term occurred at time t + 1 and t + 2 out of the number
of times the term occurred at t 1, t, t + 1, and t + 2. Fast growing
topics are those in the top 5% of the increment rate in a given year.
Glnzel and Thijs (2012) combined bibliographic coupling, co-
word analysis, and a direct citation model. First, documents are
clustered in time slices according to their cosine similarity resulting
from bibliographic coupling and textual similarity. The core clus-
ters identied through this process are next linked across different
time slices via direct citations. Emergence is then detected by iden-
tifying clusters with exceptional growth the study also considers
emerging clusters to be those that are completely new with roots
in other clusters or existing clusters exhibiting a topic shift, but this
clearly refers to the radical novelty attribute of emergence.
Overlay mapping techniques can visually reveal knowledge
areas characterised by a rapid increase in the number of documents
(publications or patents) in the global maps of science or technol-
ogy and which therefore, in comparison with other areas, may be
growing at a faster pace. (Overlay mapping can also reveal diffusion
across disciplines and technological areas.)
Other studies instead operationalised relatively fast growth by
examining the growing number of authors involved in an emerg-
ing eld over time (e.g. Guo et al., 2011; Bettencourt et al., 2008).
For example, Bettencourt et al. (2008) found that the growth of
the population of authors in a given eld tends to be relatively
well described with epidemic models that consider novel ideas as
spreading by infecting authors.
4.3. Coherence
Coherence and its persistence over time distinguish technolo-
gies that have acquired a certain identity and momentum from
those still in a state of ux and therefore not yet emerging. Coher-
ence in contemporary technologies may be detected by examining
the scientic discourse around a given emerging technology. Ini-
tially, a variety of terms may be in use and reduction in the
number of terms may signal increasing coherence. Abbreviations
or acronyms take time to appear and, when they do, signal persis-
tence; they also indicate shared interpretations and thus coherence
(Reardon, 2014). Additional signals of coherence may come from
the creation of conference sessions, tracks, dedicated conferences
11
The tf-idf (term frequency-inverse document frequency) is an indicator that
reects the importance of a word to a document in relation to a corpus. Specically,
the tf-idf is the result of the product between two indicators: the term frequency
and inverse document frequency.
1838 D. Rotolo et al. / Research Policy 44 (2015) 18271843
and subsequently from journal special issues and new specialist
journals (Leydesdorff et al., 1994). New categories in established
classication systems may also be created (Cozzens et al., 2010).
In retrospective analyses, entropy measures can be used (Watts
and Porter, 2003) as well as clustering and factor analysis of citation
and text networks. The coherence of clusters of documents or terms
can be assessed in comparison to the overall network by applying,
for example, local network density measures as well as by exam-
ining cluster persistence over time. Furukawa et al. (2015) propose
using year-to-year coherence of conference sessions to indicate
emergence. They applied co-word analysis to generate chronologi-
cal networks of conference sessions (nodes) linked by their (cosine)
similarity as based on the keywords included in the sessions papers
keywords were selected using the tf-idf indicator. Within these
networks, emerging topics are dened as sessions where previous
conferences sessions converge according to similarity.
In a similar vein, Yoon et al. (2011) developed a NLP algo-
rithm capable of identifying properties and functions in the
sentences of patent abstracts.12 The method generates an inven-
tion propertyfunction network (IPFN). Nodes in this network
represent properties and functions. A property is what a system is or
has and is expressed by using adjectives+nouns, whereas a func-
tion is what a system does and is expressed by using verbs+nouns.
Links between nodes are dened by the co-occurrence of proper-
ties and functions in patents. Emerging properties and functions
are those clustered in small and highly dense sub-networks i.e.
de facto showing a certain degree of coherence.
The approaches discussed above examine cognitive dynamics.
However, coherence can also be assessed on the basis of changes in
the social structure. In this regard, Bettencourt et al. (2009) exam-
ined the evolution of co-authorship networks at the level of the
scientist to identify network patterns associated with the emer-
gence of new scientic elds. Increasing average number of edges
per nodes (densication), stable or decreasing average path length
between two nodes (diameter), and increasing fractional count of
edges in the largest component of the considered network were
suggested as signals of emergence and specically of the topical
transition of a eld. These indicators clearly refer to increasing
connectedness of the co-authorship network, identifying emerging
communities as an indicator of emerging technology.
4.4. Prominent impact
Emerging technologies exert a prominent impact on specic
domains or more broadly on the socio-economic system by chang-
ing the composition of actors, institutions, patterns of interactions
among those, and the associated knowledge production processes.
Scientometric methods cannot identify contemporary prominent
impact due to a lack of data and the difculty in delineating the
technology in its very early stages (e.g. keywords may still be used
by groups of actors with different meanings and in different con-
texts). Mixed qualitativequantitative approaches used by Science
and Technology Studies (STS) scholars on the role of expectations
in driving technological change are of a particular relevance.13
The main argument of the sociology of expectations is that novel
12
This enables one to overcome the main limitation of co-word analysis tech-
niques, that is the need to dene an initial set of keywords before the analysis can
be performed.
13
Scientometrics can be considered as the more quantitative end of STS work. For
this reason, the distinction we make between the two traditions is not intended
to be a particularly strong one. However, it also true that there has been relatively
little interaction between scientometrics and STS since the late1980s. Each of these
tradition has its own conferences and journals, and only a handful of researchers
operate at the interface most individuals would identify themselves as either
scientometricians or STS scholars.
technologies and fundamental changes in scientic principle do not
substantively pre-exist themselves, except and only in terms of the
imaginings, expectations and visions that have shaped their poten-
tial (Borup et al., 2006, p. 285). These expectations are real-time
representations of future technological situations and capabilities
[. . .] wishful enactments of a desired future (Borup et al., 2006, p.
285) and they play a generative role by stimulating and steering
as well as coordinating actions. Evidence of this has been found
in a number of emerging elds such as gene therapy, pharma-
cogenomics, and nanotechnology (e.g. Selin, 2007; Hedgecoe and
Martin, 2003; Martin, 1999). Expectations of the performance of
novel technologies or, more generally, the ability of novel tech-
nologies to address societal problems are both important.
News articles, editorials, review and perspective articles in pro-
fessional and academic journals, vision reports and technological
roadmaps have all been used to identify statements represent-
ing multiple and potentially competing expectations surrounding a
technology (e.g. Alkemade and Suurs, 2012; van Lente and Bakker,
2010; Bakker et al., 2011). STS work has also illuminated the central
role played by hype in technology emergence. Actors who under-
stand the constitutive role of expectations have an incentive to
raise expectations in order to motivate the funding and activity
needed to realise their preferred technological future. Hype, or
over-claimed expectation, is often the result. This over-claiming
can touch most attributes of emergence and especially prominent
impact. For example, press releases prior to the launch of the Seg-
way claimed it would change walking. Similarly, in the case of
coherence, for the government to fund nanotechnology research,
they must believe nanotechnology is a thing, as opposed to a name
applied by some to a rather miscellaneous selection of materials sci-
ence research activities. Therefore, proponents have an incentive to
claim coherence where others might disagree.
These studies have been retrospective, but their data sources
are contemporaneous with technology emergence so the method
could be extended to contemporary analyses. Moreover, mapping
of expectations can be combined with scientometrics when suitable
data become available. Gustafsson et al. (2015), for example, used
technological co-classication to identify clusters of patents, the
full-text of which is subsequently analysed qualitatively to detect
guiding images or leitbild, which are generalisations shared by sev-
eral actors which guide actors towards similar objectives. Guiding
images are used to explain the dynamics of expectations.
Retrospective analyses can rely more extensively on sci-
entometrics, although this has not been done very often.
Scientometricians have mostly focused on the detection and anal-
ysis of growth and novelty, whereas impact seems to be taken for
granted. Nonetheless, scientometrics can greatly contribute to eval-
uating the impact of a potentially emerging technology. A number
of techniques can be used to produce intelligence on the emergence
process. These include the analysis of highly cited documents, of
authorship data to generate intelligence about the actors drawn
into knowledge creation processes over time (e.g. private vs. pub-
lic organisations and incumbents vs. newcomers), and of changes
in the collaboration structure as mapped with co-authorship data
(e.g. Hicks et al., 1986; Melin and Persson, 1996). Impact on knowl-
edge production processes can instead be assessed by examining
the dynamics of cognitive networks obtained from the study of the
citations or the co-occurrence of terms across a particular set of
documents.
4.5. Uncertainty and ambiguity
Emerging technologies are characterised by uncertainty in their
possible outcomes and uses, which may be unintended and unde-
sirable, as well as by ambiguity in the meanings different social
groups associate with the given technology (Stirling, 2007; Mitchel,
 D. Rotolo et al. / Research Policy 44 (2015) 18271843
2007). For analyses of contemporary emerging technologies, news
articles, editorials, review and perspective articles on professional
and academic journals can be examined to qualitatively assess the
degree of uncertainty and ambiguity associated with an emerg-
ing technology as well as to identify possible multiple visions of
the future associated with the technology. As for the evaluation of
how prominent the impact of an emerging technology will be, an
STS approach to the mapping of expectations can be used for the
assessment of uncertainty and ambiguity.
For retrospective analyses, the evaluation of uncertainty and
ambiguity remains largely unexplored in scientometric studies,
however. The few attempts made along these lines tend to overlap
with those already discussed for the evaluation of the coherence
attribute, since the main focus has been on the measurement of
the reduction of uncertainty in scientic communication rather
than on uncertainty and ambiguity associated with the potential
impact or uses of emerging technologies. For example, the creation
of a novel category (such as a new subject category in the classi-
cation of ISI WoS), in which subsequent journals associated with
the emerging technology under examination may fall, is conceived
as an indicator of increasing redundancy in the communication
process as new journals are established and achieve a critical
mass to justify the creation of a new category, the redundancy of
the communication process associated with the considered emerg-
ing technology has also increased. Increasing redundancy, in turn,
may indicate diminishing uncertainty.14 In a similar vein, Lucio-
Arias and Leydesdorff (2009) considered words in publication titles
(which are selected by authors to position knowledge claims at
a given time), cited references (which enable authors to position
knowledge claims in the existing socio-cognitive domain), and time
as key dimensions describing the scientic discourse at the research
front of a specialty. The mutual information exchanged between
these dimensions (measured in terms of Shannon entropy) is sug-
gested as an indicator of uncertainty reduction. The gap in the
assessment of uncertainty and ambiguity represents, however, an
important arena for future research, as we will discuss in the next
section.
5. Discussion
We characterised emerging technologies on the basis of ve
attributes (i) radical novelty, (ii) relatively fast growth, (iii) coher-
ence, (iv) prominent impact, and (v) uncertainty and ambiguity
and used these to develop a framework for a coherent and system-
atic operationalisation of emerging technologies. A wide variety of
scientometric methods are available to operationalise the various
attributes of emergence. Nonetheless, these are strongly dependent
on time, on the nature of the attribute, and on the data used.
Scientometric techniques are intrinsically more effective for
retrospective analyses than contemporary examinations. Time is
required before documents such as publications and patents can
be observed and techniques can be applied longitudinally. For
example, measuring growth is particularly problematic for more
contemporary analyses. Techniques using future citations are more
sensitive to this issue than methods that rely on data available when
documents are published (e.g. co-word analysis and bibliographic
coupling). Lags in database indexing may also contribute to the time
limitations of scientometric approaches.
Scientometrics is also of little use in the operationalisation of
uncertainty and ambiguity. The focus of scientometrics has been
mainly on the detection of what is emerging, rather than on cha-
racterising the potential of what is detected to be emerging. To the
14
Personal communication with Loet Leydesdorff on 2 October 2014.
1839
best of our knowledge, this area is largely unexplored. Likewise, the
methods reviewed in this paper show no explicit focus on how the
societal aspect of prominent impact can be assessed. This is some-
what surprising when one considers the extensive scientometric
work carried out for research evaluation purposes.
Furthermore, most studies have focused on publications and
patents data that are not only sensitive to time, but also provide
limited perspectives on the multifaceted phenomenon of emerg-
ing technologies. A few studies have focused on the use of news
articles and big data sources (e.g. Google Trends). These are clearly
emerging streams in scientometric and data-mining research, but
so far little attention has been paid to the use of these novel data
sources in the context of emerging technologies.
The risk that detected technological emergence may be merely
an artefact of the method used adds to these limitations. The
reviewed methodologies rely on different models, data, thresh-
olds, clustering algorithms and parameters, the selection of which
may bias the detection of emergence towards certain patterns. For
example, technological emergence is often detected with compara-
tively static analyses rather than with dynamic examinations. Data
for a given observation period are divided into time windows and
algorithms are then applied to the sample of data included in each
time window. Results may vary with the number and length of time
windows. Shorter time windows may not identify certain patterns
of emergence because they do not capture a critical mass of docu-
ments, while longer time windows may miss cases of technologies
that exhibit emerging features for a shorter period (e.g. promising
technologies that eventually do not emerge). Also, the identied
emerging technologies may be biased towards certain topics. Small
et al. (2014), for example, found that topics identied as emerging
by the combined direct citation-co-citation approach are in areas
that are more likely to offer practical outcomes than non-emerging
topics. This may suggest that such areas attract more resources,
which, in turn, may favour the recruitment of researchers (Small
et al., 2014). Yet, the identication of these emerging areas may
also be the result of the model and data used. The eld could move
forward more condently if instead of every study using a differ-
ent data set, a standard model dataset was developed to which all
techniques could be applied and the results compared (Katz, 1996).
We have argued that qualitative STS approaches can be par-
ticularly powerful for overcoming the limits of scientometrics, for
instance, in relation to prominent impact and to uncertainty and
ambiguity. For example, mapping expectations through content
analysis of news, review articles, and policy documents can provide
important insights. Because STS focuses on human agency, the
importance of expectations and visions in steering emergence as
well as the examination of niche-regime dynamics is more appar-
ent. Hence, this tradition attempts to address questions of how
emergence happens. This may favour meaningful interpretations of
scientometric data and possibly a better conceptual understanding.
Scientometrics, in turn, can bring a more robust empirical approach
to the STS research tradition, including the capability to address
measurement error by means of statistical inference as well as to
increase the generalisability of results. Few studies have followed
a combined scientometrics-STS approach. Kuusi and Meyer (2007),
for example, applied a bibliographic coupling approach to identify
clusters of patents and then to map guiding images used by differ-
ent actors to develop a consensus around the goals and directions
during different phases of development of an emerging eld.15 Yet,
there remains great potential for substantial links and a deeper syn-
thesis between the two traditions focusing on the examination of
emergence in science and technology.
15
As noted earlier in Section 4.4, a similar mixed approach has been adopted by
Gustafsson et al. (2015).
1840 D. Rotolo et al. / Research Policy 44 (2015) 18271843
The conceptualisation and operationalisation of emerging tech-
nologies offer a number of opportunities for future research. From
a conceptual point of view, more understanding of the origins of
emerging technologies is required. In the early phase of emergence
(high levels of radical novelty and of uncertainty and ambiguity,
low levels of growth, coherence, and impact) some technologies
acquire a certain momentum to become emerging (when the
levels of attributes are subject to more dramatic change), other
technologies instead arrive at the verge of becoming emergent, but
eventually not emerge at all. Funding and research programmes,
the power distribution among actors, communities of practices,
and regulations are likely to exert a signicant impact on this pro-
cess. However, more systematic research is required on the factors
that enable a technology to eventually become emergent. This also
extends to the empirical investigation of emergence. Studies often
tend to analyse emerging technologies, without comparing them
with a counter-factual sample of technologies that had the poten-
tial of becoming emergent, but eventually did not emerge. Likewise,
we have limited knowledge of the end point of the emergence pro-
cess, i.e. when emergence is over, or perhaps prematurely grinds
to a halt or reverses.
The limitations of the use of scientometrics for the operational-
isation of some the attributes also represent important avenues for
future research. In this regard, the use of novel data sources such
as publication-full-text and funding data seems particular promis-
ing. For example, publication full-text data have been mainly used
to improve the accuracy of standard scientometric approaches
(e.g. co-citation and co-word clustering) (e.g. Boyack et al., 2013;
Glenisson et al., 2005). However, the analysis of the full-text of
publications may also provide information for operationalising
the uncertainty and ambiguity attribute of emergence. Instances
of multiple and competing envisioned applications of an emerg-
ing technology may be identied in publication sections such as
the introduction and discussion, which also have the advantage
of being structured in a relatively standard manner across pub-
lications as compared to other sections. Sentiment and narrative
analysis techniques may be particularly suitable for the extraction
of this information.
Funding data may also provide relevant information for the
operationalisation of emerging technologies. For example, uncer-
tainty and ambiguity may be indicated by more extensive public
funding than private investment. Growth in funding may indicate
relatively fast growth, thus overcoming the time lag between actual
emergence and emergence detected in publications and patents.
The amount of funding can also cast light on the expected impact
of the technology. Relatively large investments suggest promi-
nent impact is expected. Nonetheless, the coverage of funding
data remains limited (Hopkins and Siepel, 2013). A number of
databases (e.g. Researchsh, FundRef, RCUK Gateway to Research,
NIH RePORTER) have been recently built with aim of providing
access to these data. Such databases include data on funding from
major funders (e.g. government departments, research councils,
large charities and foundations), but inevitably lack information on
a large variety of relatively small funding organisations that may be
important, especially in the early phases of development. The use
of funding data as reported by authors in the acknowledgements
section of publications provides better coverage of funders but no
information on the amount of funding.
The use of big data and altmetrics (e.g. download statistics,
number of retweets, Mendeley readers, citations in blogs or news
articles) add to the set of potential data sources. Given that these
data are produced in a real-time manner as compared to con-
ventional scientometric data, they seem particularly promising in
enabling the development of indicators for early detection. For
example, publication download statistics can provide an early indi-
cation of relatively fast growth and perhaps of prominent impact
in the academic domain as compared to conventional citation
data. Numbers of tweets and citations in blogs or news articles
may instead provide an indication of attention outside the aca-
demic domain. Nonetheless, there is rst a need to improve our
understanding of these data as well as how the data can be com-
pared across different cases of emerging technologies. We hope
the framework offered here can be used to structure exploration of
novel data sources for the detection of emerging technology.
6. Conclusions
Emerging technologies have assumed increasing relevance in
the context of policy-making for their perceived ability to change
the status quo (e.g. Martin, 1995; Day and Schoemaker, 2000;
Alexander et al., 2012; Cozzens et al., 2010). This has spurred ad hoc
governmental actions such as the Future & Emerging Technolo-
gies (FET) initiative funded by the European Commission in 2013
and the Foresight and Understanding from Scientic Exposition
(FUSE) research program funded by the US Intelligence Advanced
Research Projects Activities (IARPA) in 2011. The FUSE program,
for example, in pursuit of potential uses of big data, has aimed to
develop methods for the reliable early detection of emergence in
science and technology by mining the full-text of publications and
patents. Policy interest has been matched by the academic com-
munity who have developed a variety of methods for the detection
and analysis of technological emergence in recent years especially
in the scientometric domain (e.g. Small et al., 2014; Glnzel and
Thijs, 2012).
Despite this broad interest, a widely accepted denition of
emerging technologies and an agreed conceptually grounded
framework for their operationalisation are both still missing. We
showed that emerging technologies are either loosely dened in the
empirical literature or often no denition at all is provided. As a con-
sequence, operationalisations of emergence tend to differ greatly
even between approaches using the same techniques. In addition,
the understanding of what is an emerging technology differs across
actors: some individuals may conceive a technology to be emergent
because they expect impact on the socio-economic system, while
others may see the same technology as old and no longer emergent.
This, in turn, has signicant implications for policy making and the
governance of emerging technologies.
The present paper has attempted to move the eld forward
by systematically delineating the concept of technological emer-
gence linked to measurement options. To do so, we rst developed
a denition of emerging technologies that is able to capture the
multifaceted nature of emerging technologies, and then proposed a
framework for their operationalisation drawing on, but not limited
to, scientometric analysis. We identied ve attributes of emerg-
ing technologies: (i) radical novelty, (ii) relatively fast growth, (iii)
coherence, (iv) prominent impact, and (v) uncertainty and ambigu-
ity, and dened emerging technologies as: a relatively fast growing
and radically novel technology characterised by a certain degree of
coherence persisting over time and with the potential to exert a con-
siderable impact on the socio-economic domain(s) which is observed
in terms of the composition of actors, institutions and the patterns of
interactions among those, along with the associated knowledge pro-
duction processes. Its most prominent impact, however, lies in the
future and so in the emergence phase is still somewhat uncertain and
ambiguous.
We then developed a coherent and systematic framework for
operationalising these attributes of emergence. Scientometric liter-
ature was the main source of potential measures. Relevant studies
were reviewed and linked to the attributes of emergence. Our
analysis showed that scientometric analysis is particularly appro-
priate for the operationalisation of growth, novelty and coherence.
 D. Rotolo et al. / Research Policy 44 (2015) 18271843
Relatively fast growth is operationalised in many studies and often
evaluated by counting documents over time (such as news articles,
publications, and patents) (e.g. Porter and Detampel, 1995). Radical
novelty is identied with the appearance of new clusters of docu-
ments or words in citation or co-word analyses (e.g. Kajikawa and
Takeda, 2008), while other studies point to the importance of also
considering the extent to which the new cluster is connected to
clusters in the same year of observation or to clusters identied in
previous years (e.g. Small et al., 2014). Indicators based on entropy
measures or on the appearance of new categories (e.g. journals,
technological classes, terms in institutionalised vocabularies) were
identied as more suitable for assessing coherence (e.g. Cozzens
et al., 2010).
Nonetheless, important limitations exist on the scientometric
contribution to the operationalisation of emerging technologies.
The evaluation of uncertainty and ambiguity as well as prominent
impact is, for example, largely unexplored in scientometrics. Also,
methods often rely on few data sources, mostly publication and
patent data, which tend to be not suitable for the analysis of con-
temporary cases of emerging technologies these data require
time to be generated. The risk that detecting apparent emergence
may be merely an artefact of selected models adds to these limita-
tions.
We have argued that the qualitative investigation of emerg-
ing technologies conducted by STS researchers seems particularly
promising in complementing scientometrics for the purpose of
operationalising the attributes of emergence. The mapping of
expectations of emerging technologies by means of qualitative
analysis of documents such as news, review articles, and policy
documents can, for example, provide important insights on the
uncertainty and ambiguity and the prominent impact attributes
of emergence, especially in the case of contemporary analyses.
STS approaches can also provide meaningful interpretation of the
results of scientometrics, thus potentially reducing the likelihood
of detecting false positives or missing patterns.
We envisage a number of opportunities for future research.
First, future research should pay more attention to the origins of
emerging technologies. We have limited knowledge of factors that
enable certain technologies to become emergent while other do
not emerge at all. This also extends to the research design used for
the investigation of emerging technologies. Studies often examine
emerging technologies without delineating a counter-factual sam-
ple of technologies that did not emerge but which nevertheless had
the potential to emerge. Similarly, we have limited knowledge on
when a technology ceases to be emergent and what factors shape
this process. Second, the increasing access to publication full-text,
funding data, altmetrics, and, more generally big data, may provide
signicant opportunities for future research in scientometrics to
develop indicators and methods for the evaluation of attributes of
emergence for which the current state of the art in scientometrics
can provide only a limited contribution.
In summary, we have showed that considerable disagreement
exists on what is technological emergence and how it should
be operationalised. This has important implications for policy-
making in the context of emerging technologies (e.g. resource
allocation, creation of research programmes, drawing up of reg-
ulations), which, in turn, exerts a direct effect on the emergence
process itself. The present paper has attempted to contribute to
this ongoing and urgent debate in science policy research through
conceptual clarication of the phenomenon of emergence. This is a
necessary precondition for a coherent and systematic operationali-
sation of emerging technologies, for future research developments,
for a better understanding of the phenomenon, and, therefore
for more informed policy-making and governance of emerging
technologies.
1841
Acknowledgements
We acknowledge the support of the People Programme (Marie
Curie Actions) of the European Unions Seventh Framework
Programme (FP7/2007-2013) (award PIOF-GA-2012-331107
NET-GENESIS: Network Micro-Dynamics in Emerging Technolo-
gies www.danielerotolo.com/netgenesis). We are grateful to Loet
Leydesdorff, Nils Newman, Alan Porter, Andrew Stirling, Jan Youtie,
the two anonymous referees of the SPRU Working Paper Series
(SWPS), and the two anonymous referees of Research Policy for
their comments, criticisms and suggestions. A previous version of
this paper was presented at the SPRU Wednesday Seminar Series
at the University of Sussex (27 May 2015), the Technology Pol-
icy Assessment Centre (TPAC) Seminar Series at the School of
Public Policy of the Georgia Institute of Technology (4 December
2014), and the 2015 Eu-SPRI Conference (912 June 2015, Helsinki,
Finland).
Appendix
Table A1
The concept of emergence in complex systems theory (studies are ordered
chronologically).
Study Denition
Bedeau (1997) [. . .] Emergent phenomena are somehow constituted
by, and generated from, underlying processes [. . .] are
somehow autonomous from underlying processes (p.
375) [. . .] there is a system, call it S, composed out of
micro level parts [. . .] S has various macro level states
(macrostates) and various micro level states
(microstates) [. . .] there is a microdynamic, call it D,
which governs the time evolution of Ss microstates [. . .]
I dene weak emergence as follows: Macrostate P of S
with microdynamic D is weakly emergent iff P can be
derived from D and Ss external conditions but only by
simulation (pp. 377378)
Goldstein (1999) Emergence [. . .] as the arising of novel and coherent
structures, patterns, and properties during the process of
self-organization in complex systems [. . .] common
properties that identify them as emergent:
Radical novelty: emergents have features that are not
previously observed in the complex system under
observation [. . .]
Coherence or correlation: emergents appear as
integrated wholes that tend to maintain some sense of
identity over time. This coherence spans and correlates
the separate lower-level components into a higher-level
unity.
Global or macro level: [. . .] the locus of emergent
phenomena occurs at a global or macro level [. . .]
Dynamical: emergent phenomena are not pre-given
wholes but arise as a complex system evolves over time
[. . .]
Ostensive: emergents are recognized by showing
themselves, i.e. they are ostensively recognized [. . .]
(pp. 4950)
Corning (2002) Emergent phenomena be dened as a subset of the vast
(and still expanding) universe of cooperative
interactions that produce synergistic effects of various
kinds, both in nature and in human societies [. . .] all
emergent phenomena produce synergistic effects, but
many synergies do not entail emergence. In other words,
emergent effects would be associated specically with
contexts in which constituent parts with different
properties are modied, reshaped, or transformed by
their participation in the whole. (pp. 2324)
1842 D. Rotolo et al. / Research Policy 44 (2015) 18271843
Table A1 (Continued)
Study Denition
Chalmers (2006) A high-level phenomenon is strongly emergent with
respect to a low-level domain when the high-level
phenomenon arises from the low-level domain, but
truths concerning that phenomenon are not deducible
even in principle from truths in the low-level domain
[. . .] a high-level phenomenon is weakly emergent
with respect to a low-level domain when the
high-level phenomenon arises from the low-level
domain, but truths concerning that phenomenon are
unexpected given the principles governing the
low-level domain. (p. 244)
de Haan (2006) Emergence is about the properties of wholes
compared to those of their parts, about systems having
properties that their objects in isolation do not have.
Emergence is also about the interactions between the
objects that cause the coming into being of those
properties, in short the mechanisms producing
novelty. (p. 294)
Source: search performed by authors on Google Scholar and SCOPUS and extended
to cited references.
References
Abercrombie, R.K., Udoeyop, A.W., Schlicher, B.G., 2012. A study of scientometric
methods to identify emerging technologies via modeling of milestones. Scien-
tometrics 91 (2), 327342.
Adner, R., Levinthal, D., 2002. The emergence of emerging technologies. Calif. Man-
age. Rev. 45 (1), 5066.
Alexander, J., Chase, J., Newman, N., Porter, A., Roessner, J., 2012. Emergence as a con-
ceptual framework for understanding scientic and technological progress. In:
2012 Proceedings of PICMET12: Technology Management for Emerging Tech-
nologies, pp. 12861292.
Alkemade, F., Suurs, R.A., 2012. Patterns of expectations for emerging sustainable
technologies. Technol. Forecast. Soc. Change 79 (3), 448456.
Arthur, W.B., 2007. The structure of invention. Res. Policy 36 (2), 274287.
Avila-Robinson, A., Miyazaki, K., 2011. Conceptualization and operationalization of
emerging technologies: a complementing approach. In: 2011 Proceedings of
PICMET11: Technology Management in the Energy Smart World, pp. 112.
Avila-Robinson, A., Miyazaki, K., 2013a. Dynamics of scientic knowledge bases
as proxies for discerning technological emergence: the case of MEMS/NEMS
technologies. Technol. Forecast. Soc. Change 80 (6), 10711084.
Avila-Robinson, A., Miyazaki, K., 2013b. Evolutionary paths of change of emerging
nanotechnological innovation systems: the case of ZnO nanostructures. Scien-
tometrics 95 (3), 829849.
Bakker, S., van Lente, H., Meeus, M., 2011. Arenas of expectations for hydrogen
technologies. Technol. Forecast. Soc. Change 78 (1), 152162.
Bedeau, M.A., 1997. Weak emergence. Philos. Perspect. 11, 375399.
Bengisu, M., 2003. Critical and emerging technologies in materials, manufacturing,
and industrial engineering: a study for priority setting. Scientometrics 58 (3),
473487.
Bettencourt, L., Kaiser, D., Kaur, J., Castillo-Chvez, C., Wojick, D., 2008. Population
modeling of the emergence and development of scientic elds. Scientometrics
75 (3), 495518.
Bettencourt, L.M., Kaiser, D.I., Kaur, J., 2009. Scientic discovery and topological
transitions in collaboration networks. J. Informetr. 3 (3), 210221.
Boon, W., Moors, E., 2008. Exploring emerging technologies using metaphors: a
study of orphan drugs and pharmacogenomics. Soc. Sci. Med. 66 (9), 19151927.
Bornmann, L., Leydesdorff, L., 2011. Which cities produce more excellent papers
than can be expected? A new mapping approach, using Google Maps, based on
statistical signicance testing. J. Am. Soc. Inf. Sci. Technol. 62 (10), 19541962.
Borup, M., Brown, N., Konrad, K., van Lente, H., 2006. The sociology of expectations
in science and technology. Technol. Anal. Strateg. Manage. 18 (3), 285298.
Boyack, K.W., Klavans, R., Small, H., Ungar, L., 2014. Characterizing the emergence
of two nanotechnology topics using a contemporaneous global micro-model of
science. J. Eng. Technol. Manage. 32, 147159.
Boyack, K.W., Small, H., Klavans, R., 2013. Improving the accuracy of co-citation
clustering using full text. J. Am. Soc. Inf. Sci. Technol. 64 (9), 17591767.
Breitzman, A., Thomas, P., 2015. The Emerging Clusters Model: a tool for identify-
ing emerging technologies across multiple patent systems. Res. Policy 44 (1),
195205.
Callon, M., Courtial, J.J.P., Turner, W.A., Bauin, S., 1983. From translations to prob-
lematic networks: an introduction to co?word analysis. Soc. Sci. Inf. 22 (2),
191235.
Chalmers, D.J., 2006. Strong and weak emergence. In: Clayton, P., Davies, P. (Eds.), The
Re-emergence of Emergence. Oxfor University Press, Oxford, UK, pp. 244256.
Chang, C.K.N., Breitzman, A., 2009. Using patents prospectively to identify emerging,
high-impact technological clusters. Res. Eval. 18 (5), 357364.
Chen, C., 2006. CiteSpace II: detecting and visualizing emerging trends and transient
patterns in scientic literature. J. Am. Soc. Inf. Sci. Technol. 57 (3), 359377.
Cho, T.-S., Shih, H.-Y., 2011. Patent citation network analysis of core and emerging
technologies in Taiwan: 19972008. Scientometrics 89 (3), 795811.
Ciarli, T., Coad, A., Rafols, I., 2013. Quantitative analysis of technology futures: a
review of techniques, uses and characteristics. Nesta Work. Pap. 13 (8), 164.
Coccia, M., 2005. Technometrics: origins, historical evolution and new directions.
Technol. Forecast. Soc. 72 (8), 944979.
Corning, P.A., 2002. The re-emergence of emergence: a venerable concept in search
of a theory. Complexity 7 (6), 1830.
Corrocher, N., Malerba, F., Montobbio, F., 2003. The emergence of new technologies
in the ICT eld: main actors, geographical distribution and knowledge sources.
Working Papers of Faculty of Economics. Universit degli Studi dellInsubria.
Cozzens, S.E., Gatchair, S., Kang, J., Kim, K.-S., Lee, H.J., Ord nez, G., Porter, A., 2010.
Emerging technologies: quantitative identication and measurement. Technol.
Anal. Strateg. Manage. 22 (3), 361376.
Day, G.S., Schoemaker, P.J.H., 2000. Avoiding the pitfalls of emerging technologies.
Calif. Manage. Rev. 42 (2), 833.
de Haan, J., 2006. How emergence arises. Ecol. Complex. 3 (4), 293301.
de Rassenfosse, G., Dernis, H., Guellec, D., Picci, L., van Pottelsberghe de la Potterie,
B., 2013. The worldwide count of priority patents: a new indicator of inventive
activity. Res. Policy 42 (3), 720737.
De Solla Price, D., 1963. Little Science, Big Science. . .and Beyond. Columbia Univer-
sity.
rdi, P., Makovi, K., Somogyvri, Z., Strandburg, K., Tobochnik, J., Volf, P., Zalnyi, L.,
2012. Prediction of emerging technologies based on analysis of the US patent
citation network. Scientometrics 95 (1), 225242.
Fisher, J., Pry, R., 1971. A simple substitution model of technological change. Technol.
Forecast. Soc. Change 3 (1), 7588.
Furukawa, T., Mori, K., Arino, K., Hayashi, K., Shirakawa, N., 2015. Identifying the
evolutionary process of emerging technologies: a chronological network anal-
ysis of World Wide Web conference sessions. Technol. Forecast. Soc. Change 91
(February), 280294.
Gareld, E., Sher, I.H., Torpie, R.J., 1964. The Use of Citation Data in Writing the
History of Science. Institute for Scientic Information, Philadelphia.
Glnzel, W., Thijs, B., 2012. Using core documents for detecting and labelling new
emerging topics. Scientometrics 91 (2), 399416.
Glenisson, P., Glnzel, W., Persson, O., 2005. Combining full-text analysis and bib-
liometric indicators. A pilot study. Scientometrics 63 (1), 163180.
Goldstein, J., 1999. Emergence as a construct: history and issues. Emergence 1 (1),
4972.
Grupp, H., 1994. The measurement of technical performance of innovations by tech-
nometrics and its impact on established technology indicators. Res. Policy 23 (2),
175193.
Guo, H., Weingart, S., Brner, K., 2011. Mixed-indicators model for identifying
emerging research areas. Scientometrics 89 (1), 421435.
Gustafsson, R., Kuusi, O., Meyer, M., 2015. Examining open-endedness of expecta-
tions in emerging technological elds: the case of cellulosic ethanol. Technol.
Forecast. Soc. Change 91 (February), 179193.
Halaweh, M., 2013. Emerging technology: what is it? J. Technol. Manage. Innov. 8
(3), 1920.
Hedgecoe, A., Martin, P., 2003. The drugs dont work: expectations and the shaping
of pharmacogenetics. Soc. Stud. Sci. 33 (3), 327364.
Henderson, R.M., Clark, K.B., 1990. Architectural innovation: the reconguration of
existing product technologies and the failure of established rms. Adm. Sci. Q.
35 (1), 930.
Hicks, D., Martin, B.R., Irvine, J., 1986. Bibliometric techniques for monitoring perfor-
mance in technologically oriented research: the case of integrated optics. R&D
Manage. 16 (3), 211223.
Ho, J.C., Saw, E.-C., Lu, L.Y., Liu, J.S., 2014. Technological barriers and research trends in
fuel cell technologies: a citation network analysis. Technol. Forecast. Soc. Change
82 (February), 6679.
Hopkins, M.M., Siepel, J., 2013. Just how difcult can it be counting up R&D funding
for emerging technologies (and is tech mining with proxy measures going to be
any better)? Technol. Anal. Strateg. Manag. 25 (6), 655685, http://dx.doi.org/
10.1080/09537325.2013.801950
Hung, S.-C., Chu, Y.-Y., 2006. Stimulating new industries from emerging technolo-
gies: challenges for the public sector. Technovation 26 (1), 104110.
Irvine, J., Martin, B.R., 1984. Foresight in Science: Picking the Winners. Dover Frances
Pinter, London, UK.
Ivanova, V.S., Bunin, I.J., Nosenko, V.I., 1998. Fractal material science: a new direction
in materials science. JOM 50 (1), 5254.
Iwami, S., Mori, J., Sakata, I., Kajikawa, Y., 2014. Detection method of emerging
leading papers using time transition. Scientometrics 101 (2), 15151533.
Jrvenp, H.M., Mkinen, S.J., Seppnen, M., 2011. Patent and publishing activity
sequence over a technologys life cycle. Technol. Forecast. Soc. Change 78 (2),
283293.
Jun, S.-P., 2012. A comparative study of hype cycles among actors within the socio-
technical system: with a focus on the case study of hybrid cars. Technol. Forecast.
Soc. 79 (8), 14131430.
Jun, S.-P., Yeom, J., Son, J.-K., 2014. A study of the method using search trafc to
analyze new technology adoption. Technol. Forecast. Soc. Change 81 (January),
8295.
Kajikawa, Y., Takeda, Y., 2008. Structure of research on biomass and bio-fuels: a
citation-based approach. Technol. Forecast. Soc. Change 75 (9), 13491359.
Kajikawa, Y., Yoshikawa, J., Takeda, Y., Matsushima, K., 2008. Tracking emerging
technologies in energy research: toward a roadmap for sustainable energy.
Technol. Forecast. Soc. Change 75 (6), 771782.
 D. Rotolo et al. / Research Policy 44 (2015) 18271843
Katz, J.S., 1996. Bibliometric standards: personal experience and lessons learned.
Scientometrics 35 (2), 193197.
Katz, J.S., 2006. Indicators for complex innovation systems. Res. Policy 35 (7),
893909.
Kay, L., Newman, N., Youtie, J., Porter, A.L., Rafols, I., 2014. Patent overlay mapping:
visualizing technological distance. J. Assoc. Inf. Sci. Technol.
Kessler, M.M.M., 1963. Bibliographic coupling between scientic papers. Am. Doc.
14 (1), 1025.
Kleinberg, J., 2002. Bursty and hierarchical structure in streams. In: Proceedings of
the ACM SIGKDD International Conference on Knowledge Discovery and Data
Mining, pp. 91101.
Kuusi, O., Meyer, M., 2007. Anticipating technological breakthroughs: using bib-
liographic coupling to explore the nanotubes paradigm. Scientometrics 70 (3),
759777.
Lee, W.H., 2008. How to identify emerging research elds using scientometrics: an
example in the eld of information security. Scientometrics 76 (3), 503525.
Lewes, G.H., 1875. Problems of Life and Mind. Kegan Paul, Trench, Turbner, London,
UK.
Leydesdorff, L., Bornmann, L., 2012. Mapping (USPTO) patent data using overlays to
Google Maps. J. Am. Soc. Inf. Sci. Technol. 63 (7), 14421458.
Leydesdorff, L., Cozzens, S., Van den Besselaar, P., 1994. Tracking areas of strategic
importance using scientometric journal mappings. Res. Policy 23 (2), 217229.
Leydesdorff, L., Kushnir, D., Rafols, I., 2014. Interactive overlay maps for US patent
(USPTO) data based on International Patent Classication (IPC). Scientometrics
98 (3), 15831599.
Leydesdorff, L., Rafols, I., 2011. The local emergence and global diffusion of research
technologies: an exploration of patterns of network formation. J. Am. Soc. Inf.
Sci. Technol. 62 (5), 846860.
Leydesdorff, L., Rafols, I., Chaomei, C., 2013. Interactive overlays of journals and the
measurement of interdisciplinarity on the basis of aggregated journal-journal
citations. J. Am. Soc. Inf. Sci. Technol. 64 (12), 25732586.
Leydesdorff, L., Rotolo, D., Rafols, I., 2012. Bibliometric perspectives on medical inno-
vation using the medical subject Headings of PubMed. J. Am. Soc. Inf. Sci. Technol.
63 (11), 22392253.
Lin, J., 1991. Divergence measures based on the Shannon entropy. IEEE Trans. Inf.
Theory 37 (1), 145151.
Lucio-Arias, D., Leydesdorff, L., 2009. An indicator of research front activity: measur-
ing intellectual organization as uncertainty reduction in document sets. J. Am.
Soc. Inf. Sci. Technol. 60 (12), 24882498.
Manseld, E., 1961. Technical change and the rate of imitation. Econometrica 29 (4),
741766.
Martin, B.R., 1995. Foresight in science and technology. Technol. Anal. Strateg. Man-
age. 7 (2), 139168.
Martin, P.A., 1999. Genes as drugs: the social shaping of gene therapy and the recon-
struction of genetic disease. Sociol. Health Illn. 21 (5), 517538.
Melin, G., Persson, O., 1996. Studying research collaboration using co-authorships.
Scientometrics 36 (3), 363377.
Mitchel, S., 2007. The import of uncertainty. Pluralist 2 (1), 5871.
Moed, H.F., 2010. Measuring contextual citation impact of scientic journals. J.
Informetr. 4 (3), 265277.
Morris, A.S., Yen, G., Wu, Z., Asnake, B., 2003. Timeline visualization of research
fronts. J. Am. Soc. Inf. Sci. Technol. 54 (5), 413422.
Narin, F., Hamilton, K.S., Olivastro, D., 1997. The increasing linkage between U.S.
technology and public science. Res. Policy 26 (3), 317330.
Nightingale, P., 2014. What is technology? Six denitions and two pathologies, SPRU
Working Paper Series, 2014-19:1-29.
Ohniwa, R.L., Hibino, A., Takeyasu, K., 2010. Trends in research foci in life science
elds over the last 30 years monitored by emerging topics. Scientometrics 85
(1), 111127.
Ottino, J.M., 2004. Engineering complex systems. Nature 427 (6973), 399.
Porter, A.L., Ashton, W.B., Clar, G., Coates, J.F., Cuhls, K., Cunningham, S.W., Ducatel, K.,
van der Duin, P., Georgehiou, L., Gordon, T., Linstone, H., Marchau, V., Massari, G.,
Miles, I., Mogee, M., Salo, A., Scapolo, F., Thissen, W., 2004. Technology futures
analysis: toward integration of the eld and new methods. Technol. Forecast.
Soc. 71 (3), 287303.
Porter, A.L., Detampel, M.J., 1995. Technology opportunities analysis. Technol. Fore-
cast. Soc. Change 49 (3), 237255.
1843
Porter, A.L., Roessner, J.D., Jin, X.-Y., Newman, N.C., 2002. Measuring national emerg-
ing technology capabilities. Sci. Public Policy 29 (3), 189200.
Rafols, I., Porter, A.L., Leydesdorff, L., 2010. Science overlay maps: a new tool for
research policy and library management. J. Am. Soc. Inf. Sci. Technol. 61 (9),
18711887.
Reardon, S., 2014. Text-mining offers clues to success. Nature 509 (7501), 410.
Roche, I., Besagni, D., Francois, C., Hrlesberger, M., Schiebel, E., 2010. Identication
and characterisation of technological topics in the eld of molecular biology.
Scientometrics 82 (3), 663676.
Rogers, E., 1962. The Diffusion of Innovations. Free Press of Glencoe, New York.
Rotolo, D., Rafols, I., Hopkins, M., Leydesdorff, L., 2014. Scientometric mappings as
strategic intelligence for tentative governance of emerging science and tech-
nologies, SPRU Working Paper Series, 2014-10:1-40.
Sahal, D., 1985. Technological guideposts and innovation avenues. Res. Policy 14 (2),
6182.
Saviotti, P., Metcalfe, J., 1984. A theoretical approach to the construction of techno-
logical output indicators. Res. Policy 13 (3), 141151.
Scharnhorst, A., Gareld, E., 2010. Tracing scientic inuence. Dyn. Socio-Econ. Syst.
2 (1), 133.
Schiebel, E., Hrlesberger, M., Roche, I., Francois, C., Besagni, D., 2010. An advanced
diffusion model to identify emergent research issues: the case of optoelectronic
devices. Scientometrics 83 (3), 765781.
Selin, C., 2007. Expectations and the emergence of nanotechnology. Sci. Technol.
Hum. Values 32 (2), 196220.
Shibata, N., Kajikawa, Y., Takeda, Y., Sakata, I., Matsushima, K., 2011. Detecting
emerging research fronts in regenerative medicine by the citation network anal-
ysis of scientic publications. Technol. Forecast. Soc. Change 78 (2), 274282.
Small, H., 1973. Co-citation in the scientic literature: a new measure of the rela-
tionship between two documents. J. Am. Soc. Inf. Sci. 24 (4), 265269.
Small, H., 2006. Tracking and predicting growth areas in science. Scientometrics 68
(3), 595610.
Small, H., Boyack, K.W., Klavans, R., 2014. Identifying emerging topics in science
and technology. Research Policy 48 (8), 14501467, http://dx.doi.org/10.1016/
j.respol.2014.02.005
Srinivasan, R., 2008. Sources, characteristics and effects of emerging technologies:
research opportunities in innovation. Ind. Mark. Manage. 37 (6), 633640.
Stahl, B.C., 2011. What does the future hold? A critical view on emerging information
and communication technologies and their social consequences. In: Chiasson,
M., Henfridsson, O., Karsten, H., DeGross, J.I. (Eds.), Researching the Future in
Information Systems: IFIP WG 8.2 Working Conference, Future IS 2011. Turku,
Finland, June 68, 2011. Proceedings. Springer, Heidelberg, pp. 5976.
Stirling, A., 2007. Risk, precaution and science: towards a more constructive pol-
icy debate. Talking point on the precautionary principle. EMBO Rep. 8 (4),
309315.
Takeda, Y., Kajikawa, Y., 2008. Optics: a bibliometric approach to detect emerging
research domains and intellectual bases. Scientometrics 78 (3), 543558.
van Lente, H., Bakker, S., 2010. Competing expectations: the case of hydrogen storage
technologies. Technol. Anal. Strateg. Manage. 22 (6), 693709.
van Lente, H., Rip, A., 1998. The rise of membrane technology. From rethorics to
social reality. Soc. Stud. Sci. 28 (2), 221254.
Vincenti, W.G., 1984. Technological knowledge without science: the innovation of
ush riveting in American airplanes, ca. 1930ca. 1950. Technol. Cult. 25 (3),
540576.
Wassermann, S., Faust, K., 1994. Social Network Analysis: Methods and Applications.
Cambridge University Press, Cambridge, MA.
Watts, R.J., Porter, A.L., 2003. R&D cluster quality measures and technology maturity.
Technol. Forecast. Soc. Change 70 (8), 735758.
Yan, E., 2014. Research dynamics: measuring the continuity and popularity of
research topics. J. Informetr. 8 (1), 98110.
Yoon, J., Choi, S., Kim, K., 2011. Invention property-function network analysis of
patents: a case of silicon-based thin lm solar cells. Scientometrics 86 (3),
687703.
Zhang, Y., Porter, A.L., Hu, Z., Guo, Y., Newman, N.C., 2014. Term clumping for tech-
nical intelligence: a case study on dye-sensitized solar cells. Technol. Forecast.
Soc. Change 85 (June), 2639.